Project Reference No.: UGC/FDS16/P03/25
Project Title: Topological Insulators-Based Surfaces as a Novel Platform for Enhanced Surface Plasmon Resonance Sensing
Principal Investigator: Dr ASSI Dani Samer (HKMU)
Abstract
Enhancing surface plasmon resonance (SPR) with wavelength-selective tunability has potential to transform the development of plasmonic devices to unleash the immense potential of plasmonic technology in sensing, biomedical diagnostics and future environmental monitoring applications. However, conventional predefined periodic metals with high conductivities, such as gold and silver, which are widely regarded as traditional plasmonic materials, suffer from significant energy dissipation due to interband electronic transitions and Drude losses, limiting their efficiency and sensitivity in plasmonic applications. These materials also face additional drawbacks and challenges in practical device implementation and integration, including limited operational bandwidth, poor tunability, complex fabrication processes and most critically high costs, which make them impractical for large-scale or cost-sensitive applications.
This has inspired a search for novel low-loss, low-cost and highly tunable plasmonic materials to enable the current technology to achieve enhanced efficiency and broader operational bandwidth - capabilities that cannot be guaranteed by conventional plasmonic materials. In this regard, we aim to design and develop tunable topological insulator (TI)-based surfaces, including Tin-based chalcogenides (SnSe, SnTe, SnSeTe), Antimony-based chalcogenides (SbSe, SbTe, SbSeTe), and Bismuth-based chalcogenides (BiSe, BiTe, BiSeTe), with properties optimized to enhance surface plasmon resonance and enable tunable selective spectral responses.
Topological insulator material exhibits strong spin-orbit coupling, leading to Dirac surface states that are protected from backscattering, allowing charge carriers from these surface states to carry current and move freely parallel to the surface. These properties enable enhanced coupling between surface electrons and surface plasmon modes across a broad spectral range (ultraviolet to terahertz), enabling diverse plasmonic applications.
In this project, the proposed approach of tuning (Se) and (Te) ratios would influence the energy levels, charge carrier dynamics and the spin-orbital coupling (interlocking of spin state and charge state), thereby maximizing interaction with surface plasmons while minimizing plasmonic losses. Therefore, tuning the (Se) and (Te) ratios is essential for enhancing SPR effectiveness and tuning the spectral bandwidth in the proposed TI-based surfaces.
Experimentally, we will utilize an electrochemical synthesis technique to fabricate these TI-based surfaces, ensuring precise control over their elemental composition and structure. Our recent publications on the electrochemical synthesis of topological insulator materials (Dani S. Assi et al., Advanced Materials 2024, 36, 2306254 & Dani S. Assi et al, Advanced Science 2023,10, 2300791) and preliminary simulation results have proven the feasibility of the proposed work.
In this project, we challenge the current limitations of traditional plasmonic materials by exploring the potential of tunable TI-based surfaces to enhance SPR effects, reduce plasmonic losses, and extend selective response capabilities for versatile next-generation sensing applications. We will take a systematic approach to investigate the material composition and structural characterization procedures for the proposed TI-based surfaces.
The key deliverables from our proposed work include: (1) Synthesis of topological insulator materials using electrochemical deposition technique, with a focus on optimizing their nanostructures; (2) Fabrication of wafer-scale (4-inch) tunable TI-based surfaces; (3) Theoretical study and experimental characterization of the surface plasmons and electronic properties of the developed TI surfaces; and (4) A TI-based sensing platform with enhanced SPR efficacy and tunable spectral range for detecting the fingerprints of three different antibiotics.
The novelty of this proposal is producing tunable TI-based films under simple fabrication procedure that could enable higher efficiency with selective spectral responses for practical sensing applications, while significantly reducing costs compared to traditional plasmonic materials. Successful execution of this work will advance plasmonic technology in sensing, diagnostics, and environmental monitoring, contributing to Hong Kong and the Greater Bay Region, a major technology hub with high population density.
Project Reference No.: UGC/FDS16/E07/25
Project Title: Quality Enhancement for MPEG Video-based Point Cloud Compression (V-PCC)
Principal Investigator: Dr AU YEUNG Siu-kei (HKMU)
Abstract
In recent years, there’s been growing interest in using 3D video and image data in many areas, including self-driving cars, creating digital replicas of cities, virtual reality and augmented reality (VR/AR), and preserving cultural heritage. One popular type of 3D data is called a point cloud, which is highly valued for its ability to represent detailed and complex scenes.
A point cloud is a collection of individual points that represent the surfaces of objects in 3D space. Each point in the cloud contains geometric information, such as its 3D coordinates, and can also include additional details like surface orientation, brightness, and color.
Point clouds, which are collections of data points in space, can be categorized based on their application into point cloud scenes and point cloud objects. Point cloud scenes are typically captured in real-time by LIDAR sensors, such as those mounted on vehicles for self-driving navigation. On the other hand, point cloud objects can be either static or dynamic. Dynamic point clouds, which change over time, are used in technologies like augmented reality (AR), virtual reality (VR), volumetric video, and telepresence. Each frame of a point cloud can contain millions of data points, leading to massive amounts of data. For instance, a standard dynamic point cloud dataset can have around one million points per frame, with each point requiring about 60 bits of data. This means that an uncompressed dynamic point cloud running at 30 frames per second can generate around 180 megabytes of data every second. Transmitting or storing such large volumes of data demands significant bandwidth and storage capacity, making it challenging to use in practical applications.
To effectively reduce the size of point cloud data, the Moving Picture Experts Group (MPEG) has developed two standard compression methods. The first method, called Geometry-based Point Cloud Compression (G-PCC), uses a technique that organizes data into a tree-like structure. This method is particularly effective for compressing static point cloud scenes. The second method, known as Video-based Point Cloud Compression (V-PCC), employs a technique that projects data into smaller, manageable patches. This approach is designed for compressing dynamic point cloud objects that change over time.
Under the V-PCC standard, a point cloud is first converted from its 3D form into a series of 2D patches, which are then organized into a conventional 2D video sequence. Traditional video compression methods, such H.264 and H.265, can be used to compress this sequence. However, this process can significantly degrade the quality of the reconstructed 3D point cloud. The degradation occurs due to errors introduced during the conversion between 2D and 3D, as well as from the compression itself. These errors are especially problematic during low-bit-rate transmission. The compression errors can generally be classified into three types: (1) Geometric Misalignment: The positions of the original data points shift to different locations after compression. (2) Attribute Distortion: The original color or intensity of the data points changes after compression. (3) Data Point Loss: Some data points are missing after compression.
Addressing these compression errors is particularly challenging because point clouds are a relatively new type of data representation. Most existing techniques for enhancing the quality of digital images and videos do not work well with these sparse, three-dimensional datasets. In this research project, we aim to develop a comprehensive solution to improve the quality of compressed point cloud data. Our solution will focus on jointly optimizing the point cloud quality by addressing geometric misalignment, attribute distortion, and data point loss.
Given the complexity of the compression process, our enhancement operations can be applied both during the compression (within the compression loop) and after the compression (outside the compression loop). We hope that our newly designed enhancement scheme will eventually become a valuable option for the next generation of MPEG V-PCC or MPEG PCC standards. Additionally, we plan to explore whether our proposed scheme can also be applied to other point cloud compression methods, such as G-PCC and end-to-end compression.
Project Reference No.: UGC/FDS15/H29/25
Project Title: Remembering through Scanning: Exploring the Affordances for Spatial Memory Practices in Mobile 3D Scanning
Principal Investigator: Dr CAI Shengdan (Shue Yan)
Abstract
Mobile 3D scanning technologies have recently become more accessible, enabling everyday users to capture spaces as 3D models with ease. While initially adopted for professional purposes such as design, documentation, and real estate, mobile 3D scanning is now being used to preserve spaces of personal significance, such as childhood homes or transitional living spaces. This trend is particularly relevant in the context of increasingly mobile populations (e.g., digital nomads, expatriates, and international students). As mobile 3D scanning becomes more popular, it shows potential to become an integral part of memory-making, much like mobile photography. Despite its growing use, the implications of mobile 3D scanning for memory studies remain underexplored. Research on digital memory has often focused on themes of connectivity and placelessness. While these characteristics are praised for liberating memory from physical boundaries, they are also criticised for fostering shallow and fleeting engagements with places. A tension emerges in digital memory studies between the static, fixed nature of physical space and the fluid, dynamic flow of information within connected digital networks. However, the rise of locative media challenges this perceived incompatibility. For example, Augmented Reality applications enable tourists to intensively interact with both superimposed digital content and physical sites. Mobile 3D scanning pushes this shift further by compelling users to digitally preserve spaces they find meaningful, while also prompting them to revisit these spaces, along with the memories associated with them, through the act of scanning. Investigating the relationship between mobile 3D scanning and memory practices offers an opportunity to bridge this theoretical incompatibility by integrating human actors, digital tools, and ongoing memory practices within the same physical space.
While existing theories of memory media often emphasise representational outcomes such as the final 3D models produced, this study shifts the focus to the ongoing memory practices that unfold during the process of scanning. Unlike the instantaneous act of photography, scanning requires users to physically move through and interact with a space, enabling them to revisit familiar places, uncover forgotten corners, and notice overlooked objects. As users navigate their surroundings with the scanning device, their attention, perceptions, and actions evolve, transforming memory-making into an active and embodied process. To investigate this ongoing process, this study draws on the concept of affordance, which highlights the mutuality of an actor’s perceptions and the capabilities of objects, environments, or technologies. This framework allows for an examination of the dynamic interactions between human actors, digital tools, and physical spaces, focusing on how these interactions afford specific memory practices. This study will employ a participatory workshop approach in which participants will be trained in mobile 3D scanning techniques and tasked with documenting spaces of personal significance. Through follow-up reflections and discussions, participants will provide insights into what memory practices are afforded during pre-scanning preparation, the scanning process, and post-scanning review. By tapping into mobile 3D scanning as an entry point, this study reconciles the incompatibility between physical and digital by taking into account the interplay between human actors, digital tools, and physical space in affording memory practices. Beyond academic contributions, the findings have practical applications for improving public literacy in 3D technologies and guiding the design of future tools for memory preservation. While the study focuses on the production of memory, it also lays the foundation for future research into how 3D-scanned memories are shared and experienced by others.
Project Reference No.: UGC/FDS16/E15/25
Project Title: Bearing-type bolted connections of wire arc additively manufactured steel under ambient, fire and post-fire conditions
Principal Investigator: Dr CAI Yancheng (HKMU)
Abstract
Additive Manufacturing (AM) has already been applied in various engineering disciplines, including aerospace, mining equipment and automotive. It is being explored in the construction industry in recent years due to its great potential including structural design of steel structures. Wire Arc Additive Manufacturing (WAAM) is widely recognized as the most suitable AM technique for steel structures due to its relatively high deposition rate, large build volume, and low cost. This digital fabrication technology is currently available in China and other regions. WAAM offers new opportunities to address challenges in the construction industry, such as material waste, the shortage and aging of skilled workers in Hong Kong and other areas, and safety issues on construction sites. It also provides benefits such as optimized structural geometries and ease of production in remote locations. These advantages align with the public policy goals of the Government of the Hong Kong Special Administrative Region (HKSAR), which include construction waste reduction, energy efficiency, and sustainable development. The WAAM also aligns with the “Construction 2.0” published by the Development Bureau of HKSAR, which promotes the use of advanced technologies in the construction industry, with a particular focus on enhancing site safety and creativity. However, the steel structures fabricated using the WAAM technique are relatively new to engineers and researchers. Currently, there is no international design code for WAAM steel structures or their structural components due to limited investigations. This lack of standardized guidelines restricts the broader adoption of this technique in the construction sector.
Bolting is one of the most commonly used connection types in the assembly of steel structural components and units. It facilitates load transfer between structural steel elements such as braces, beams, and columns, as well as between structural steel units like modules. The load transfer mechanisms in these connections are typically categorized as either bearing-type (where the bolt bears against the connected components) or friction-type (where friction between the connected components is achieved by tightening the bolts to a specified pretension force), or a combination of both. The structural behaviour of bolted connections is crucial for ensuring the integrity and stability of structures under various conditions, including ambient, fire, and post-fire scenarios. Despite the significant interest and need for adopting the WAAM technique in the steel construction industry, there are currently no established design rules for bolted connections of WAAM steel under ambient, fire, and post-fire conditions. Additionally, there are few studies in the literature that investigate bolted connections of WAAM steel, particularly under fire and post-fire conditions.
This research project proposes to investigate the structural behaviour of bearing-type bolted connections in WAAM steel under ambient, fire, and post-fire conditions. Fire and post-fire conditions will involve temperatures up to 1000 °C. WAAM steel connection specimens will be fabricated using feedstock wires of varying strengths, specifically ER70S and ER120S, with nominal yield strengths up to 830 MPa. Different printing directions will also be considered in the study. Both experimental and numerical investigations will be conducted to assess the structural performance of bearing-type bolted connections in WAAM steel. The objectives of this project are to generate experimental and numerical data on the resistances, bolt-hole deformations, and failure modes of these connections under ambient, fire, and post-fire conditions. Additionally, the project aims to derive new design rules for different failure modes to facilitate the adoption of WAAM steel in the construction industry. This research will contribute to more innovative, efficient, and reliable designs for steel structures using the WAAM technique.
Project Reference No.: UGC/FDS51/H01/25
Project Title: Vowel Length Distinction in Old Chinese as Reflected by the Prosody of the Book of Odes
Principal Investigator: Dr CHAN Abraham Yee-shun (UOWCHK)
Abstract
The Book of Odes, aka Shijing, is a collection of classical Chinese poetry allegedly edited by Confucian himself. These poems were composed, sung and recited in Old Chinese, a term that refers specifically to the pronunciation of the Chinese language spoken before and during the Han dynasty.
While Qing dynasty literati realized that Old Chinese was very different from the language they were familiar with, and made tremendous effort to study it, the first systematic reconstruction of this lost sound of the pre-Han era was due to the eminent Swedish sinologist Bernhard Karlgren. But Karlgren did not really start from scratch—his Old Chinese reconstruction was essentially a projection of his Early Middle Chinese (Qieyun) reconstruction on top of the Qing studies of Old Chinese rhyming and phonetic loans. Karlgren hypothesized a medial glide /j/ to account for the Type A/B distinction in Early Middle Chinese, and he projected this glide all the way to Old Chinese.
Many linguists have since tried to eliminate this glide from their own Old Chinese reconstructions as it has made their reconstructions very unnatural. One solution is to postulate a vowel length distinction in Old Chinese and assume that the glide emerged either from long or short vowels. But if this feature really existed in Old Chinese, one should expect it to have an effect on the prosody of Shijing, yet I have demonstrated that this cannot be further from the truth.
The present project aims to put forward my own reconstruction of Old Chinese finals, which I also hypothesize to possess a vowel length distinction. But the reason behind my decision is entirely different. In an earlier study I have eliminated Karlgren’s medial glide from my Early Middle Chinese reconstruction by positing a vowel system with tense and lax vowels, and I now need to explain how this vowel system evolved from Old Chinese. My proposed solution suggests that, parallel to the history of Latin and many other languages, Old Chinese had long and short vowels, which later developed into tense and lax counterparts. A major vowel shift, perhaps not unlike the one that affected English, disoriented the post-Han rhyming practice and gave us Early Middle Chinese. Vowel length in my Old Chinese reconstruction, therefore, does not directly correspond to the Type A/B distinction.
A brief test demonstrates that my proposed vowel length distinction coincides with Shijing meters. For more than a millennium, Chinese scholars unanimously agreed that classical poetry before the time of Shen Yue, in contrast to the regulated verse popularized during the Tang dynasty, only had meters based on number of syllables. My theory, however, suggests that the Book of Odes already possessed quantitative meters based on vowel length, just like poetry of some other classical languages such as Latin, Greek and Sanskrit. If proven correct, it will revolutionize our understanding of Old Chinese, the nature, content and transmission of Shijing, and the history of Chinese literature.
Project Reference No.: UGC/FDS24/B21/25
Project Title: Technostress and Experience Economy: The Effects of Augmented and Virtual Reality in Museums from Different Generational Cohorts
Principal Investigator: Dr CHAN Elaine Ah-heung (PolyU SPEED)
Abstract
In the era of virtual tourism and service automation, immersive technologies such as Virtual Reality (VR) and Augmented Reality (AR) are increasingly utilized to provide aspirational tourism experiences. Research indicates that VR and AR significantly enrich visitors’ experiences in museums through immersive and interactive elements, which traditional exhibits may lack. These technologies profoundly influence the tourism industry, impacting various stages of the tourist decision-making process. By 2024, VR and AR are projected to contribute to a $72.8 billion global market. Specifically, VR and AR enhance visualization of destinations and attractions, provide previews of hotels and increase temptation to visit by offering a 'try-before-you-buy' experience to reduce perceived risk. At tourist destinations, VR and AR offer extra sensory experiences through aesthetic, entertainment, and educational aspects, as well as digital guiding and navigation, and they increase tourist engagement and satisfaction. VR and AR can also encourage tourists to revisit destinations virtually. However, the global aging population presents challenges, as senior travelers, who are not as techno-oriented, may experience technostress, potentially deteriorating their visit experiences. Senior travelers are expected to become the largest potential market for international tourism, with significant growth projected in the coming years. By 2030, 1 in 6 people will be aged 60 years or over, and by 2050, the population of people aged 60 years and older will reach 2.1 billion. In 2024, travelers aged 60 years and above accounted for 37% of total travelers, up from 16.45% in 2020-2021. Senior travelers travel an average of 27 days per year, and more than 52% consider travel and vacation a priority for discretionary income. While digital technologies are embedded in many tourism and entertainment operations to enhance effectiveness, VR and AR can also create cognitive load and distraction. Therefore, this research aims to develop an integrated model to understand senior tourist’s' attitudes and behavior towards VR and AR tourism experiences in the museum context. The specific research objectives are to develop an integrated model to examine how VR and AR experiences affect tourists' satisfaction and repeat visit intention in the museum context, identify the association between museum experience economy using VR and AR and the customer perceived value of their VR and AR museum experience, examine the relationship between customer perceived value and customers’ behavioral outcomes, discuss the moderated role of 'technostress' and 'age' in the formation of customers' satisfaction and memory of experience, as well as explore differing perceptions and attitudes towards VR and AR tourism experiences among different generational cohorts. Mixed methods will be employed to study how the different generational cohorts view the role of VR/AR in influencing their behavioral intentions in terms of memory of experience and visit satisfaction. The findings from this study may provide valuable insights for destination management organizations (DMOs), tourism attractions, museums, travel agencies, and other relevant parties to formulate appropriate digital strategies for the travelers' markets to enhance the tourism experience, provide better visitor satisfaction and encourage more repeat patronage for the attractions. By doing so, tourists will then have a higher sense of loyalty of the destination.
Project Reference No.: UGC/FDS24/E16/25
Project Title: Development of a Cloud-based Fuzzy Assessment Framework for Smart Age Friendly Housing for Middle Class Elderly in Hong Kong
Principal Investigator: Dr CHAN Joseph Hing-lun (PolyU SPEED)
Abstract
Population ageing is a global trend that has garnered attention from policy makers and academia. Hong Kong, like other developed economies, has seen a rise in the proportion of elderly persons aged 65 and over in the total population. It is projected that this percentage will increase to 32.3% in 2039. As a highly urbanised city, ageing in Hong Kong has ramifications that extend across different disciplines including real estate, engineering, architecture, and social science. The development of age-friendly cities has become a fertile research area attracting much attention from researchers in the field of ageing and built environment. To tackle the problems associated with ageing populations such as changes in population structure, pension shortages, and increasing medical expenses, ageing in place (AIP) is advocated to achieve successful ageing. Successful ageing refers to low probability of diseases, good cognitive and functional capacity, and active engagement in the lives of the elderly.
Although AIP has been promoted in different countries around the globe, one of the major challenges is that many senior citizens’ home conditions are less than satisfactory to support their later lives. Many homes in Hong Kong were built with outdated building standards in the last century, while age-friendliness of housing may not be a prime concern of developers and individuals. Previous research in AIP in Hong Kong mainly focused on low-income class whose housing needs are usually addressed by public rental housing usually. Compared with low-income groups, middle class elderly are relatively disadvantaged in the sense that they are not eligible for various government subsidies in housing, social welfare and outreach support service to elderly.
This research aims to develop a cloud-based computerized framework for assessing the age friendliness of housing for middle class in Hong Kong, incorporating digital inclusion and smart technology. The resulting Overall Age Friendliness Index will aid in decision-making for building design and maintenance of housing for the elderly. Additionally, a user-friendly, cloud-based computerized system will be created to visualize and compare the age friendliness of different housing projects in Hong Kong.
In terms of short-term impacts, the index would raise awareness about the importance of age-friendly housing and encourage individuals, policymakers, and developers to evaluate the existing housing options. It would help identify gaps and areas for improvement in the current housing stock. In addition, the evaluation process would lead to immediate improvements in the living conditions of older adults. Issues such as accessibility, safety, and comfort can be addressed promptly, ensuring that middle class older residents have suitable and comfortable housing options. For medium-to-long impacts, the index can serve as a benchmark to establish minimum standards for age-friendly housing. This could lead to the development of regulations and guidelines that ensure new housing projects cater to the needs of older adults, promoting accessibility, adaptability, and safety. Age-friendly housing can enhance social inclusion by creating communities where older adults can live independently and actively participate in society. The index can drive the development of housing that fosters social connections, promotes intergenerational interaction, and supports aging in place.
Project Reference No.: UGC/FDS16/M06/25
Project Title: A study on the toxicological responses of Scenedesmus species to different sizes of micro/nano-plastics
Principal Investigator: Dr CHAN Sidney Man-ngai (HKMU)
Abstract
Plastics have become integral to modern life, yet their lifecycle often results in fragmentation into microplastics (MPs), ranging from 1 μm to 5 mm, and nanoplastics (NPs), those below 1 μm. The ubiquitous presence of micro/nano-plastics (MNPs) in different compartments of environments, including air, water, soil, sediment and biota, raising concerns about their toxicity to aquatic organisms across various trophic levels. These pollutants can disrupt gastrointestinal function, impair growth, induce immunotoxicity, and bioaccumulate within food webs, classifying them as emerging pollutants requiring urgent investigation.
Microalgae, as primary producers, are essential to ecosystem functioning and the global carbon cycle. Alterations in their physiology due to MNP exposure can disrupt food chains by reducing primary productivity and nutrient cycling, destabilize ecosystems by diminishing biodiversity and resilience, and lead to long-term ecological consequences. These cascading effects underscore the critical importance of understanding the impacts of MNPs on microalgae and aquatic ecosystems. Despite existing studies reporting both positive and negative effects of MNPs on microalgae, the results remain controversial due to differences in experimental conditions, species-specific responses, and the types and characteristics of MNPs used. While some studies have focused on well-studied species such as Chlorella and Synechococcus, the responses of other widely distributed microalgae, including Scenedesmus, remain less understood. Moreover, the underlying molecular responses, such as transcriptomic changes, remain poorly characterised.
This project aims to address this knowledge gap by investigating the physiological and molecular-level responses of two Scenedesmus isolates from a local estuary, Scenedesmus quadricauda and Scenedesmus acuminatus, to varying polymer types, sizes and concentrations of MNPs. Specific objectives include i) examine the physiological responses of the two Scenedesmus isolates to MNPs of varying types, sizes and concentrations, to determine how these factors influence the overall health of the microalgae; ii) compare the biochemical defensive mechanisms employed by the two Scenedesmus isolates in response to MNP exposure; and iii) elucidate the molecular mechanisms underlying the responses of the two Scenedesmus isolates to MNPs, to provide insights into the gene expression changes that enable this genus to cope with environmental stressors.. This research will enhance our understanding of microalgal responses to MNPs, contributing valuable insights into the ecological impacts of MNPs pollution in aquatic environments. Ultimately, this knowledge will inform risk assessments and the development of mitigation strategies to protect aquatic ecosystems from the adverse effects of plastic pollution.
Project Reference No.: UGC/FDS11/M03/25
Project Title: Roles of the amygdala in GLP-1 receptor agonist-induced nausea and emesis
Principal Investigator: Dr CHAN Sze-wa (SFU)
Abstract
Obesity and diabetes have become critical public health problems affecting 8 – 10% population globally. Glucagon-like peptide-1 (GLP-1) receptor agonists represent a novel class of medications to treat type 2 diabetes and obesity. However, 40 – 70% patients may develop adverse gastrointestinal side effects, including nausea and emesis, limiting the dose that can be used. Understanding the mechanism of the side effects associated with GLP-1 receptor agonists is important for developing new treatments with improved tolerability. However, most studies have focused on the brainstem, and this limits our understanding of mechanisms involved in nausea.
There is evidence that nausea and emesis induced by GLP-1 receptor agonists involves an activation of emesis pathways in the brain. We previously found that GLP-1 receptors in the brainstem and hypothalamus may be involved in exendin-4- and GLP-1 (7-36) amide-induced emesis in Suncus murinus. Central administration of these drugs also increases c-Fos expression in the central nucleus of the amygdala (CeA) where GLP-1 receptors are located. By using radiotelemetry techniques, we found that exendin-4 induces behaviour indicative of nausea. Recently, we showed that peripherally administered exendin-4 induces emesis and increases c-Fos expression in the CeA, nucleus tractus solitarius, and area postrema. Intracerebral paraventricular hypothalamic (PVH) administration of baclofen antagonizes emesis induced by peripherally administered exendin-4, without modulating its actions to inhibit food intake, suggesting a role of hypothalamic γ-aminobutyric acid B (GABAB) receptors in exendin-4-induced emesis. Furthermore, continuous infusion of ghrelin into the CeA antagonizes cisplatin-induced emesis while des-acyl-ghrelin enhances the cardiovascular and respiratory changes caused by cisplatin. The CeA receives projections from the parabrachial nucleus and PVH and sends descending projections to the brainstem to modulate emotions and mediate the perception of nausea. We hypothesize that the CeA may be involved in the mechanism of nausea and emesis of GLP-1 receptor agonists. Activation of the inhibitory GABAB neurons in the CeA may attenuate signal transduction pathway of GLP-1 receptors, ultimately alleviating nausea and/or emesis induced by GLP-1 receptor agonists.
In this project, animal experiments will be conducted using a combination of behavioral testing and established radiotelemetry recording techniques to assess physiological changes indicative of nausea, coupled with c-Fos immunohistochemical and western blot analysis of brain functions. Overall, these studies will provide biological and molecular insights pertaining to GLP-1 receptor agonists-induced side effects which may ultimately help guiding developing of improved anti-diabetic and anti-obesity therapies. Our new knowledge may also uncover novel mechanisms of nausea and emesis leading to advanced therapeutics for anti-emetic that would have a major public health impact.
Project Reference No.: UGC/FDS11/H02/25
Project Title: Hong Kong Southbound Writer during the Cultural Cold War: Xu Xu and the Development of “Visual Novel” Games
Principal Investigator: Dr CHAN Winnie Hiu-ting (SFU)
Abstract
From 1950 to 1969, the Chinese writer Xu Xu (1908-1980) published numerous novels in Hong Kong newspapers and periodicals such as Sing Tao Wan Pao and World Today, which effectively demonstrated his dedication to literary writing. Yet despite Xu’s prominence as a Hong Kong literary figure and representative of the “southbound intellectuals”, his works have not been adequately examined in relation to contemporary socio-political dynamics in Hong Kong.
To address this important gap, the proposed project will first recontextualize Xu’s overlooked Hong Kong novels within the framework of the Cultural Cold War and “Greenback culture”. This will involve an in-depth examination of the correlation between Xu’s works and local politics. The second part of the project will go beyond traditional textual analysis in humanities research by reconstructing and restoring the reading experiences of Xu’s contemporary readers. The third part of the project aims to achieve knowledge transfer by bridging literary research with artificial intelligence and design, creating a series of “visual novel” games based on Xu’s novels.
By making Xu’s works accessible through this interactive platform, the project will not only contribute to the emerging field of “techno-humanities” but also foster a renewed interest in modern Chinese literature, particularly among younger audiences who are more accustomed to digital media than traditional formats.
Project Reference No.: UGC/FDS14/H06/25
Project Title: Transformation and Reconstruction: A Study of the Reception of Qu Yuan and the Chuci during the Republican China Era
Principal Investigator: Dr CHEN Hung-to (HSUHK)
Abstract
During the Republic of China (ROC, 1912-1949), China underwent significant cultural transformations, showcasing the intellectual fusion of Western and Eastern culture and serving as an intermediary in the transition from tradition to modernity. In response to the impact of Western academic thought, ROC intellectuals reexamined Chinese native culture and innovated upon the foundation of traditional scholarship, thus creating a vibrant academic landscape. Scholars of the ROC dedicated considerable efforts to the study of Qu Yuan and the Chuci (Songs of the South), including notable figures such as Hu Shi (1891-1962), Liang Qichao (1873-1929), Gu Jiegang (1893-1980), Fu Sinian (1896-1950), Guo Moruo (1892-1978), and Wen Yiduo (1899-1946). According to preliminary statistics from this project, there are now over a hundred published works on ROC Chuci studies, spanning a wide range of fields, including literature, philosophy, geography, folklore, and mythology. The research methods employed were influenced by Western scholarship while retaining traditional academic characteristics. Therefore, it can be said that the study of Chuci in the ROC reflects the diverse characteristics of an era marked by tremendous change.
This project argues that the Chuci scholarship of the ROC represents an important transitional phase from traditional to modern Chuci studies and should be established as a distinct area of research to explore its historical and academic significance in depth. This project aims to focus on how Qu Yuan and the Chuci were received during the ROC period, particularly how Qu Yuan’s image evolved under the influence of Western thought. The project will elaborate on the transformation of traditional Chuci studies in light of the influx of Western learning and assess the impact and significance of Western knowledge on Chuci scholarship, as well as the insights it brings to modern Chuci studies. Ultimately, through interdisciplinary and multi-methodological approaches, this project aims to apply its findings to the study of other Chinese traditional literature and textual studies, thereby offering new insights into pre-modern Chinese literature and intellectual history.
Project Reference No.: UGC/FDS16/P05/25
Project Title: Laser-Enabled Surface Engineering of PDMS: Advancing Modular Microfluidic Platforms with Enhanced Antifouling and Electrochemical Sensing Capabilities
Principal Investigator: Dr CHEN Jianlin (HKMU)
Abstract
Polydimethylsiloxane (PDMS) is a widely used material for fabricating microfluidic devices, but its surface properties often need to be modified to enhance microfluidic devices’ performance. While laser induced graphene (LIG) technology has emerged as promising approach for surface engineering, current LIG-based methods for PDMS modification have limitations, such as the need for carbon-rich additives (e.g. polyimide powder) or the requirement to transfer LIG structures (e.g. from polyimide film) onto the PDMS surface.
To address these limitations, this project proposes a novel, direct approach to engineer the PDMS surface using carbon dioxide (CO2) laser technology, without the need for additional carbon-rich additives. The key objectives of this study are to: 1) optimize the laser treatment parameters to precisely control the characteristics of modified PDMS surface, 2) investigate the mechanism and kinetics of CO2 laser-induced surface modifications on PDMS, and 3) integrate the modified PDMS surfaces into microfluidic systems and electrochemical sensors.
Unlike previous methods relying on graphene-like structures, this proposed direct CO2 laser approach offers a more streamlined and customizable pathway to impart desirable properties, such as enhanced electrochemical performance and improved antifouling behavior, directly on the PDMS surface. By systematically varying the laser parameters, including velocity, power, frequency, and pulse density, the research team will establish a comprehensive understanding of how to tailor the PDMS surface modifications to meet the specific requirements of microfluidic and electrochemical sensing applications.
Furthermore, the project will investigate the underlying mechanism and kinetics of the CO2 laser-induced surface modifications on PDMS through advanced characterization techniques. The modified PDMS surfaces will then be integrated into microfluidic systems to evaluate their antifouling performance and used as electrode materials in modularly customizable electrochemical sensors, assessing their sensitivity, selectivity, and stability for various environmental monitoring applications, such as the detection of organic compounds, and microbial contaminants.
The outcomes of this research have the potential to significantly benefit various stakeholders, including the microfluidics and biosensing industries, as well as environmental monitoring agencies and authorities. By expanding the understanding and application of direct laser-based surface engineering of PDMS, this project will contribute to the development of advanced, multifunctional materials and devices that can enhance environmental monitoring capabilities and support sustainable solutions. The anticipated impact of this research aligns with Hong Kong's innovation and technology development blueprint, as outlined in the country's 14th Five-Year Plan for National Economic and Social Development and Vision 2035.
Project Reference No.: UGC/FDS16/E16/25
Project Title: Development of Concept-based Visual State Space Model for Interpretable Ulcerative Colitis Diagnosis
Principal Investigator: Dr CHEN Xueli (HKMU)
Abstract
Ulcerative colitis (UC) is a chronic inflammatory bowel disease marked by continuous mucosal inflammation and ulceration starting in the rectum and potentially extending throughout the colon. UC significantly poses a significant burden on patients’ quality of life, affecting physical health, psychological well-being, social interactions, and professional life. Accurate assessment of UC severity is essential for therapeutic decisions and monitoring disease progression. The Mayo endoscopic score (MES) is widely used in developing automatic UC severity scoring systems. Standardizing UC evaluation through the MES system enables meaningful comparisons across studies and facilitates precise, personalized treatment strategies.
Current automated UC severity scoring methods, such as CNNs and Vision Transformers, face two key limitations: (1) difficulty in efficiently modeling long-range dependencies in endoscopic images, and (2) lack of interpretability, which hinders their adoption in clinical settings. Recent advancements in State Space Models, particularly the Mamba architecture, offer advantages with linear computational complexity, hardware-efficient implementation, and the ability to model long-range dependencies while maintaining a global effective receptive field. Our research group has recently developed a Mamba-based model, UCMamba, for UC severity scoring. While UCMamba has demonstrated its effectiveness in our preliminary experiments, there is still room for improvement in its ability to preserve spatial relationships for local and global information. Therefore, one goal of this project is to propose image-level and patch-level contrastive learning to improve the representation learned by UCMamba. Our proposed image-level contrastive learning captures better global information while patch-level contrastive learning captures fine-grained local information.
Interpretability is another key focus of our work. In the healthcare field, it is essential to have models that are both high-performing and understandable, which drives the need for research into explainable artificial intelligence. Recently, ante-hoc explainable methods, particularly concept-based methods, have attracted great interest. The MES system assesses inflammation in UC on a 4-point scale ranging from 0 (normal) to 3 (severe), based on levels of erythema, vascular patterns, friability, and ulceration. Given the semantic nature of MES, concept-based methods offer a promising solution for UC diagnosis. Therefore, another goal of this project is to develop an inherently interpretable UC diagnosis model using concept-based methods. Specifically, our proposed framework seeks to project UCMamba’s feature space into a space where the axes correspond to interpretable concepts, and predictions are made based on the weighted average of these concepts. This framework will provide interpretations of model predictions in both visual and textual formats. For example, the attention mechanism in UCMamba can be visualized to highlight the important regions that contribute to the final predictions. Additionally, the weights of the linear predictor indicate the importance of each concept to the final decision, allowing the concept bottleneck layer to elucidate the reasoning behind individual decisions.
Our framework has the potential to revolutionize UC severity assessment by providing an accurate, automated, and interpretable scoring system. By bridging the gap between machine learning and clinical practice, UCMamba will empower gastroenterologists to make more informed therapeutic decisions, ultimately improving patient outcomes.
Project Reference No.: UGC/FDS16/H45/25
Project Title: Rethinking Incentivized Review Regulations: A Stakeholder-Centered Approach to Enhancing E-Commerce Transparency in China
Principal Investigator: Dr JI Li (HKMU)
Abstract
Online reviews play a crucial role in shaping consumer decisions, but their reliability is increasingly threatened by incentivized reviews—where sellers offer rewards in exchange for positive feedback. In China, this issue is particularly widespread, leading to misleading information, unfair competition, and declining consumer trust. In response, the Chinese government implemented a nationwide ban on incentivized reviews in 2024. However, despite this prohibition, non-compliance remains a persistent challenge, as businesses find ways to evade detection, platforms struggle with enforcement, and consumers continue to encounter biased reviews.
This research seeks to understand why the ban has not been fully effective and to explore alternative regulatory approaches to improve compliance and restore trust in online reviews. The study takes a multi-perspective approach, examining the views of three key stakeholder groups: consumers, businesses, and e-commerce platforms. Through interviews and case studies, it investigates how consumers perceive online reviews, whether they can identify biased content, and how the ban has influenced their trust in review authenticity. It also explores how businesses interpret and comply with the ban, the challenges they face in enforcement, and their perspectives on potential regulatory improvements. Additionally, it examines how e-commerce platforms detect and prevent incentivized reviews, the difficulties they encounter in enforcement, and their views on alternative compliance strategies.
Beyond China, many countries use different regulatory approaches to address the problem of fake or biased reviews. Some governments emphasize strict enforcement, while others focus on transparency-based measures, such as requiring businesses to disclose incentivized reviews. This research conducts a comparative legal analysis of regulatory models from the UK, EU, USA, and Australia to assess whether elements of these frameworks could be adapted to China’s legal and economic environment. By identifying best practices and evaluating their feasibility in China, the study provides practical recommendations for improving regulatory enforcement, enhancing business accountability, and empowering consumers to make informed purchasing decisions.
The findings of this research will contribute to both academic and policy discussions on consumer protection, e-commerce regulation, and digital platform governance. The study will propose concrete policy recommendations, such as mandatory disclosure rules, stricter platform accountability measures, and consumer education initiatives, to create a more transparent and trustworthy online marketplace. Ultimately, this project aims to develop a regulatory framework that can effectively combat misleading reviews, protect consumer rights, and promote ethical business practices, not only in China but also as a reference for other digital economies facing similar challenges.
Project Reference No.: UGC/FDS11/H05/25
Project Title: A study of intergenerational solidarity and social exclusion of South Asian older people
Principal Investigator: Dr CHEUNG Ada Pui-ling (SFU)
Abstract
Hong Kong is a relatively homogeneous society, with ethnic Chinese comprising around 91.6% of the total population. Among the ethnic minorities, Indians, Nepalese, and Pakistanis are the three major groups. Statistical data reveals a 93.8% increase in the number of older people aged 65 or above in these three groups from 2011 to 2021. This aging trend highlights the need to pay attention to their experience of growing old in a different cultural context. The literature review shows that ethnic minority older people have suffered from exclusion in various dimensions, leading to isolation, loneliness, and unequal access to social or healthcare services. Although there are ample global studies on ethnic minorities in later life, this issue is under researched in Hong Kong. This is largely due to the assumption that South Asian older people are well looked after by their families and do not require much external support. Families are expected to play a key role in supporting them financially and emotionally, assisting with service applications, and most importantly, helping them integrate into society and reduce the impact of social exclusion on their everyday lives. Given the changing demographic and socio-economic conditions, it is unclear whether intergenerational support can continue to fulfill the traditional roles, despite the cultural emphasis in South Asian communities on children revering their older family members. This study seeks to address this knowledge and service gap by exploring the experiences of social exclusion among South Asian older people, their views on the family's role in helping them cope with exclusion, the gaps between the expected and actual roles of the family in their care, and the challenges faced by their adult children in providing family care. The study will also provide policy recommendations and service delivery alternatives. To collect in-depth information on this topic, a qualitative approach will be used, based on semi-structured interviews with 45 South Asian older people and 45 adult children of the older interviewees from three South Asian ethnic groups – Pakistani, Nepalese, and Indian. This research will focus on older people who are 60 or over and live with or apart from their children in the community. Since this research is grounded in the perspectives of both older and younger generations, it can give a voice to these ‘seldom heard’ groups regarding policies and services related to their well-being.
Project Reference No.: UGC/FDS16/H29/25
Project Title: From Exploration to Enhancement: A Mixed-Methods Study of Self-Regulation in Generative AI-Supported Academic Writing in Hong Kong
Principal Investigator: Dr CHUNG Hiu-yui (HKMU)
Abstract
Although academic writing is essential in higher education, many students, particularly learners of English as a second or foreign language, find it very challenging. This difficulty stems from the significant intellectual and linguistic demands of academic writing, with research indicating that students often struggle with generating ideas, structuring their writing, and adhering to academic conventions. While researchers have emphasised the importance of self-regulated learning (SRL) strategies to improve writing, there is insufficient research investigating strategy interventions in the context of academic writing. Meanwhile, the rapid development and increasing accessibility of generative AI (GenAI) present significant opportunities for enhancing students’ academic writing performance, although the integration of GenAI into academic contexts also poses challenges such as overreliance, misinformation, and inconsistency. To harness the potential benefits of GenAI while addressing any associated problems, it is essential to help students cultivate SRL strategies that empower them to use GenAI effectively throughout the writing process.
Accordingly, the proposed study aims to explore how students use GenAI for academic writing while identifying ways to support its use through a mixed-methods, two-phase design. In the first phase, we will develop and initially validate a questionnaire with 200 students to examine their SRL strategies and GenAI usage. Following validation, we will administer this questionnaire to another group of 200 undergraduates across various Hong Kong universities and analyse the data. Employing stratified sampling, we will then select 60 students for an academic writing task using GenAI, with analyses focusing on transcribed Zoom session recordings that capture interactions with GenAI. A subset of 12 students will be chosen for stimulated recall interviews. In the second phase, we will develop GenWISE, an intervention programme focused on GenAI for Writing Improvement and Self-Regulation Enhancement, based on the findings from the first phase and our expertise in English for Academic Purposes (EAP) teaching and GenAI. The programme will engage approximately 60 students in a six-stage pedagogical framework—planning, prompting, previewing, producing, peer-reviewing, and portfolio-tracking—to enhance their critical thinking and SRL strategies in using GenAI for academic writing. We will assess the programme’s impact by analysing changes in students’ (1) use of SRL strategies for GenAI use, (2) interactions with GenAI prompts, (3) writing outcomes and (4) critical perceptions of AI-supported writing. An inductive approach will help us investigate how different factors, such as AI literacy and reflexivity, mediate these effects. Data from multiple sources, including questionnaire responses, academic essays, follow-up interviews, GenAI interactions, and reflective writing, will undergo quantitative and qualitative analyses to provide a comprehensive understanding of the data.
The contributions of the proposed study are threefold. Theoretically, it will enhance scholarly understanding of the SRL strategies employed by undergraduates in Hong Kong when using GenAI to improve their academic writing performance. It will also illuminate whether and how strategy-based instruction, guidance on engineering prompts, and opportunities for guided self-reflection can foster changes in students’ SRL practices and academic writing proficiency. Methodologically, the study will develop and validate a survey on students’ SRL strategies with GenAI tools. We will also employ a diverse range of new data collection methods (e.g., GenAI interaction data) to gain rich insights into student engagement and provide a solid foundation for future research. Pedagogically, the study will shape EAP curricula by providing insights for instructors in Hong Kong and beyond on leveraging GenAI and promoting SRL to enhance students’ academic writing. It will also guide teachers across disciplines to effectively integrate GenAI into their teaching and inform policy-making decisions regarding its educational use.
Project Reference No.: UGC/FDS25/B03/25
Project Title: Product innovation from user-generated content analytics with traditional and generative AI: the application of induction-deduction framework
Principal Investigator: Dr CUI Xiling (THEi)
Abstract
It is common knowledge that firms are provided with richer external sources for their product innovation, namely online social media reviews. Information, using experiences and ideas about products, are shared online by many consumers. With the constant evolution of artificial intelligence (AI) technologies, firms can then turn these consumers’ online reviews into a valuable asset for co-creation innovation with consumers to extract innovative ideas for new product development.
In the past few years, scholars have tried to use different data analytical technologies to analyze online reviews for product innovation. However, the existing literature lacks a framework to guide the different practices, given that products are diversified (they are in different stages of the life cycle and may have different innovation strategies), the data amounts to be acquired from consumers are different, and there are many data analytics technologies to be used. Therefore, this project proposes to develop a comprehensive product innovation analytics framework to contribute to both the practical and academic fields.
In this project, we first review the product innovation (with two innovation strategies) and its relationship with product life cycle and customer involved co-creation innovation. Then we focus on the application of AI in user-generated content analytics for product innovation. The commonly used technologies in user-generated content analytics are reviewed and we then introduce two solutions with traditional AI and generative AI (GenAI) for product innovation. More importantly, we introduce the induction-deduction framework (IDF) as our theoretical ground, based on which, this project explores matching different products in different life cycle stages (i.e., introduction, growth, maturity, and decline) and different strategies (i.e., radical and incremental innovation), different amounts of data acquisition and accumulation, and different data analytics techniques.
This project is expected to have a significant impact from both the academic and practical perspectives. The introduction of the AI technology into product innovation analytics field is expected to not only enrich the literature body, but also inspire the academia to leverage the existing knowledge from different perspectives.
In addition, the project is expected to further enrich the literature by developing a comprehensive product innovation framework with AI technologies. It is also the first attempt at mapping several dimensions, including the knowledge of products and innovations from operations management, data analytics skills from data sciences, and the IDF from philosophy into one operational framework. This project is thus expected to inspire more interdisciplinary research.
From a practical perspective, this project is expected to provide a framework for product managers and R&D staff to effectively conduct innovation analytics. They could quickly position their product(s) and then identify effective strategies and plans for their product innovation within the framework. It is also expected that the investigators will develop and enhance data analytics skills and gain some key insights from the project for teaching AI courses.
Project Reference No.: UGC/FDS16/M10/25
Project Title: Mechanistic Insights into Microalgal-Bacterial Biofilm Interactions: Enhancing Pharmaceutical Removal from Wastewater
Principal Investigator: Dr DENG Dan (HKMU)
Abstract
Pharmaceutical pollution in wastewater has emerged as a critical environmental and public health issue, particularly in densely populated regions such as Hong Kong. Rapid urbanization and high pharmaceutical consumption rates have led to the continuous release of chiral compounds, such as atenolol, metoprolol, venlafaxine, and chloramphenicol, into municipal wastewater systems. Conventional domestic wastewater treatment plants often lack advanced processes to effectively remove these contaminants. As a result, pharmaceuticals persist in wastewater, accumulate in aquatic ecosystems, and cause significant ecological and human health concerns, including endocrine disruption, promotion of antibiotic resistance, and threats to ecosystem integrity.
Biofilms, complex microbial communities composed of bacteria, fungi and microalgae, offer a promising solution to these challenges due to their energy efficiency, reduced sludge production and synergistic pollutant degradation pathways. These consortia perform critical functions such as nutrient cycling, cross-feeding, and enhanced degradation, often driven by the activity of keystone species. However, significant knowledge gaps persist regarding the fundamental interactions within biofilms that are crucial for pharmaceutical remediation. Specifically, the mechanisms underlying nutrient exchange, signaling pathways, and cooperative behaviors remain poorly understood. Additionally, the role of keystone species in maintaining biofilm stability under pharmaceutical stress is largely unexplored. Besides, pharmaceuticals can destabilize microbial communities by disrupting cooperative interactions and altering biofilm compositions, highlighting the urgent need for further research into these dynamics.
This study aims to develop and optimize a robust microalgal-bacterial biofilm system for pharmaceutical remediation through a three-task framework focusing on biofilm formation, enhancement, and functional analysis under pharmaceutical stress. Task I will construct moving bed biofilm reactors (MBBRs) to monitor biofilm dynamics, removal efficiencies, and microbial community shifts using influent from wastewater treatment plants under winter and summer conditions. Periodic analyses of biofilm samples will assess their physicochemical properties and microbial compositions to identify functional taxa and keystone species critical for pharmaceutical degradation. Task II will focus on optimizing contaminant removal efficiency through bioaugmentation strategies. High-efficiency degraders and dominant species isolated from the biofilms established in Task I will be introduced into the system. Keystone species will also be targeted for enhancement through the supplementation of quorum sensing (QS) molecules to stimulate cooperative behaviours. These augmented strains will be integrated into MBBRs to evaluate their impact on pharmaceutical removal efficiency, biofilm stability, and microbial community structure. This approach aims to determine whether combining highly efficient degraders and dominant species with QS-mediated stimulation of keystone species can significantly improve system performance. Finally, Task III will utilize multi-omics approaches, including metagenomics and transcriptomics, to uncover the metabolic pathways and interspecies interactions driving pharmaceutical degradation. By comparing native and bioaugmented biofilms with or without additional QS molecules, this task will elucidate molecular mechanisms governing biofilm structure, removal efficiency, and microbial tolerance to pharmaceutical stress. Changes in gene expression related to degradation enzymes, EPS biosynthesis pathways, and QS-regulated signaling will be quantified. Additionally, metabolic pathways responsible for degrading chiral pharmaceuticals will be mapped alongside their transformation products.
This research will integrate biofilm engineering, microbial ecology, and advanced analytical techniques to propose a scalable, sustainable solution for pharmaceutical removal in wastewater treatment. The findings will guide future design and operation of biofilm-based systems to reduce environmental pollution, protect public health, and ensure the long-term viability of aquatic ecosystems.
Project Reference No.: UGC/FDS24/E01/25
Project Title: Study of the Distortion of Surface Acoustic Wave by Nanoparticles in Microfluidics Towards the Development of an Ultrafine Particle Sensor
Principal Investigator: Dr FU Sau-chung (PolyU SPEED)
Abstract
Particulate matter (PM) is a mix of particles suspended in air and is associated with adverse health effects. Epidemiological studies have already indicated that ambient PM2.5 (diameters less than 2.5 µm) is carcinogenic. The finer (smaller) a particle is, the deeper it penetrates the human respiratory system. Therefore, there has been growing concern about health hazards of ultrafine particles (UFPs), which are defined as particles with diameters less than 100 nm, that their adverse health effects will be even worse. However, published regulatory occupational exposure limits for airborne exposures of such nano-sized materials are still absent. Although Artificial Intelligence plus Internet of Things (AIoT) has empowered data analytics for finding correlations, epidemiological evidence for UFPs is still inconclusive because of two deficiencies. First, the PM exposure has been monitored by stationary equipment in fixed locations and is not necessarily typical of the air breathed in by people in general. Second, some portable PM sensors have been developed enabling a more realistic personal exposure, but those portable devices can only broadly measure PM2.5. Equipment for measuring UFPs is either bulky or contains a hazardous (e.g. radioactive) component. Therefore, the development of a miniature and safe UFP measurement device is essential. Recently, surface acoustic wave (SAW) devices have been actively investigated because of their high sensitivity, simple and low-cost fabrication procedure, and most importantly, their compact size. In an SAW device, two sets of electrodes are fabricated on a piezoelectric material. A sinusoidal electricity signal is applied to one set of electrodes to excite an SAW through the vibration of the piezoelectric material. Then, the wave propagates to another electrode, but it will be distorted if there is disturbance along propagation surface, e.g. the existence of UFPs. By pumping the ambient air through microfluidics and analyzing the distorted wave signal, an SAW device can be developed into a UFP sensor. To develop this technology, understanding how the particles affect the propagation of SAW is crucial. This project will first study the effects of UFPs on distortion of the SAW including wave speed, phase shift and attenuation, under different wave frequencies. Then, an SAW device will be fabricated and examined. By using the correlation data between the characteristics of the wave and particle size, with the help of a machine learning algorithm, the device will be optimized for UFP detection. Low-energy sound fields, which are harmless to humans, are used in the device, so this proposed project would make actual personal monitoring of UFPs more feasible, and consequently, stronger epidemiological associations between UFP exposure and health responses could be confirmed. The findings will determine a new direction for the development of miniature sensors so as to enable the deployment of data analytics and AIoT in the field of air quality.
Project Reference No.: UGC/FDS16/E06/25
Project Title: A Study on Inference Service-aware Edge Computing Systems: Sustainability and Human-Centric Design
Principal Investigator: Dr FU Yaru (HKMU)
Abstract
Nowadays, with the rapid advancement of wireless technologies, edge computing has emerged as a highly effective paradigm for deploying a variety of innovative applications at the network edge, which has greatly improved the accessibility of inference services for mobile users. Inference services, often supported by deep learning-based intelligence models, are performed at edge servers to enable efficient data processing. These services empower various intelligent featured apps such as image-based object recognition, natural language processing, or video-based behavior prediction. By distributing these inference jobs at network edges, systems can provide users with faster and more context-aware services.
While the seamless delivery of inference services on edge networks is promising, it requires tackling challenges that need innovative solutions in resource management to ensure smooth and scalable user experience. Moreover, as the demand for computing grows, so do electricity consumption and greenhouse gas emissions. This poses acute sustainability challenges, as in the future intelligent era, edge computing is expected to be energy-efficient to prevent excessive power use, which is particularly important for mobile and remote edge devices. The geo-distributed nature of edge servers offers a unique opportunity to address this issue by utilizing on-site renewable energy, such as solar and wind energy, making the usage of mixed energy sources a promising solution for future edge computing paradigms.
Besides, as the user base and service complexity expand, edge infrastructures must be able to adaptively manage the complex system and network behaviors, heterogeneous resource, and uncertain end-user demands. This is especially crucial when integrating energy considerations, such as the availability of green energy, multiple energy providers, and grid electricity prices and policies. Simply maximizing green energy usage may not always lead to long-term cost savings. Additionally, understanding human behavior in relation to services is another critical issue as it significantly impacts the decisions of service providers in effective service management. This influence, in turn, affects energy scheduling, resource allocation, and policy making, as different inference services typically require varying amounts of storage and energy resources, along with diverse service-level objectives (SLOs) such as accuracy, response latency, etc. However, gaining this information is challenging. On the one hand, user behavior is closely linked to factors such as time and mobility. On the other hand, an individual’s request patterns are strongly influenced by the usage behaviors of other users, a phenomenon known as the network effect or network externality in marketing and economics.
In this project, we will first explore human behavior design through the lens of the network effect. This model will consider how users, especially those acquiring the same service, influence an individual’s choice. Based on the established model, we will investigate the service recommendation problem by selecting an offer set that maximizes service provider’s profit. Subsequently, we will develop adaptive energy and service management strategies aimed at reducing carbon emissions within the networks, considering practical constraints such as SLOs and resource availability. Finally, we will examine how the reshaping of user behavior interacts with various energy, service, and resource management techniques, as these factors collectively play an essential role in enhancing the system’s sustainability and user’s satisfaction.
In final, we aim to develop a novel edge computing system prototype integrated with innovative solutions to show practical applicability for real-world entities focused on achieving green computing objectives. This prototype has the potential to benefit both network operators and mobile users. Through this project, network operators can expect lower overall operational costs, contributing to advanced decarbonization in Hong Kong’s ICT sector. While mobile users can access intelligent services empowered by virtual and physical resource management without the constraints of time and location. In addition, service providers in Hong Kong can enhance their profitability through service recommendations based on human-centric design.
Project Reference No.: UGC/FDS13/H06/25
Project Title: Ulugh Beg’s Astronomical Achievements and His Legacy in Ming China
Principal Investigator: Prof FUNG Kam-wing (Chu Hai)
Abstract
This research project aims to shed light on the life and astronomical achievements of Ulugh Beg (1394-1449, r. 1447-1449), ruler of the Timurid Empire (1370–1507) and an astronomer, and his astronomical legacy in China. Grandson of the Central Asia conqueror Tamerlane (1336-1405), he built the Ulugh Beg Observatory in Samarkand (in today’s Uzbekistan) between 1424 and 1429 before ascending to the throne and the observatory rose to become one of the most important astronomical centres in the Islamic world. With the data collected at the observatory and from other sources, he compiled the star catalogue Zij-i-Sultani in 1437 with 994 stars, a groundbreaking contribution before the Scientific Revolution of Europe.
This project is divided into two parts, each of them focusing on a distinct aspect of the astronomer's life and work. Part one investigates the life and astronomical contributions of Ulugh Beg. It aims to uncover 1) the life of Ulugh Beg before and after he became the Timurid sultan and 2) the achievements of this Central Asian astronomer, including his contributions to the field of astronomy, particularly in the realm of Islamic constellations and observational data in his star catalogue. It examines the relationship between the observatory led by the astronomer and its role in advancing astronomical knowledge for the empire, thus revealing the substantial contributions of Islamic science to the early modern world.
Part two of the project seeks to establish interconnections between Islamic and Chinese astronomy. It aims to reveal the continuous exchange of scientific knowledge between China and the other part of the world based on Ulugh Beg's star catalogue. The astronomy of the early part of the Ming Dynasty (1368–1644), especially the scientific knowledge of the government’s Astronomical Bureau, was heavily influenced by Islamic astronomy which was a legacy from the Mongolian Yuan court (1271–1368): Islamic instruments were used by the Islamic division of the Astronomical Bureau while some Islamic astronomical works were translated into Chinese during the reign of Emperor Hongwu (1328–1398, r. 1368–1398), founder of the Ming Dynasty. Comparing the scientific knowledge in Ulugh Beg’s catalogue and in astronomical works of the Ming Dynasty, the project will trace the connections between the Central Asian astronomer and the astronomy of China in the East, highlighting potential knowledge exchanges and mutual influences between these two regions.
Studies on Islamic astronomy in China written in Chinese, Japanese and Western languages are fairly limited in number and primarily focus on the Mongolian Yuan period. Early studies of Islamic astronomy in Ming China tend to focus on the institutions of astronomy. While subsequent researchers addressed the translation of Islamic astronomical texts in early Ming, this project will study an Islamic text entering China later to pinpoint that Islamic astronomy in Ming China was more than a Mongolian legacy: it was a living tradition with knowledge exchange with the Islamic world.
This research aims to enrich our understanding of the connectivity of astronomical knowledge across Central Asia, the Islamic world and China during a pivotal historical period. By examining the life and works of Ulugh Beg, the project aims to illuminate the cultural, scientific and intellectual exchanges that shaped the development of astronomy in these regions. Furthermore, this study proposes to contribute to broader discussions on the global history of science and the transmission of knowledge across civilizations of the Eurasian continent.
Keywords: Ulugh Beg, Islamic Science, History of Astronomy, Ming China, Silk Road
Project Reference No.: UGC/FDS16/P01/25
Project Title: Emerging Pollutants in Mangroves: Occurrence of Liquid Crystal Monomers (LCMs) and Their Interaction With Microplastics (MPs)
Principal Investigator: Dr HAN Jie (HKMU)
Abstract
As electronic devices such as smartphones, televisions, and computers become increasingly prevalent in our daily lives, the rising production and disposal of these devices contribute significantly to the accumulation of liquid crystal monomers (LCMs) in the environment. LCMs have emerged as pollutants linked to liquid crystal display (LCD) technology and electronic waste (e-waste). Reports indicate that millions of devices are discarded annually, leading to the release of hazardous materials, including LCMs, into ecosystems. Given the persistent, bioaccumulative, and toxic (PBT) characteristics of LCMs, global attention towards them is increasing. This underscores that LCM contamination is not merely a localized issue but a global concern that necessitates comprehensive research and mitigation efforts. LCMs can accumulate in living organisms and adsorb onto organic particles, such as microplastics (MPs), due to their lipophilicity and hydrophobicity, raising concerns about their potential classification as persistent organic pollutants (POPs). Although research on LCM contamination is still in its infancy, recent studies have detected LCMs in various environmental matrices, such as residential dust, water and sediments. This has raised further concerns about their impacts on health and ecosystems, including mangroves. In China, LCM contamination has been found in the Pearl River Estuary (PRE), a densely populated and industrially active area surrounded by mangroves. Mangrove ecosystems are essential for sediment accretion, biodiversity, protection against extreme weather, coastal erosion prevention and carbon fixation. However, they face significant threats by pollution from both ocean and land sources, especially from MPs, which are prevalent in the Pearl River Delta (PRD) region. Mangrove plants can trap pollutants like MPs and LCMs in their dense root systems, acting as natural filters and aiding pollution control. Interestingly, mangroves often exhibit higher levels of MP contamination than other coastal areas due to their unique role as physical and biogeochemical barriers that impede the movement of contaminants across intertidal zones. However, there are currently no reports indicating whether LCM contamination in mangroves is higher than in other coastal areas. Following the detection of LCMs and MPs in the PRD region, it is important to consider how MPs may interact with LCMs, as these interactions could influence the uptake of LCMs by mangrove plants. Understanding these interactions is crucial, yet the study of MP and LCM interactions in mangroves and how LCM-MP co-contamination affects the uptake of LCMs by plants remains unreported. This research project aims to fill this gap by focusing on three main parts: (1) Investigate LCM contamination: We will investigate concentration levels, types and detection frequencies of LCMs and the abundance of MPs in water, sediment and plant samples within the PRD area; (2) Study interaction mechanism with MPs and influencing factors: We will examine the interaction mechanisms and influencing factors between dominant LCMs and MPs. Sorption kinetics and isothermal analysis will be conducted under various influencing factors; (3) Assess the impact of LCM and MP on the growth of mangrove plant (Kandelia obovate, KO) at the seedling stage: We will evaluate the individual effect of LCMs and the combined effect of LCMs with MPs on the growth of a model mangrove species, KO, at the seedling stage, the most sensitive stage to pollutants. This research will be the first to explore the occurrence of LCMs in mangrove ecosystems, focusing on their interactions with MPs and the effects on mangrove plant growth. This proposed project will provide fundamental insights for understanding the characteristics of LCMs as emerging pollutants, their interaction mechanism with MPs, and their effects on the growth of mangrove plants, guiding responsible electronics and plastic products consumption practices and effective management and conservation strategies for mangrove ecosystems in China and around the world.
Project Reference No.: UGC/FDS11/E05/25
Project Title: Harnessing Large Language Models for Social Trust: Automated Fact-checking for Public Health Informatics by Large Language Models
Principal Investigator: Dr HANG Ching-nam (SFU)
Abstract
In the recent digital environment, the rapid spread of false information on online social media platforms presents a significant challenge. The World Health Organization described the health misinformation during the COVID-19 pandemic as an “infodemic”, highlighting its global scale and impact. At the same time, the emergence of large language models (LLMs) has introduced concerns about hallucination issues, further complicating the problem of digital information inaccuracies. While social media serves as a vital tool for information sharing, it often accelerates the spread of rumors, conspiracy theories, and pseudoscience. The consequences are far-reaching, affecting individuals, society, and nations through financial losses, social unrest, and poor decision-making. The complexity and scale of this problem highlight the inadequacy of our current capabilities to effectively identify misinformation. In response, several organizations and government agencies have taken steps to address the problem. For example, the Semantic Forensics program launched by Defense Advanced Research Projects Agency (DARPA) in 2021 aims to develop technologies to detect and classify fake media. Similarly, recent initiative of the United States Food and Drug Administration to create a Rumor Control hub represents a proactive approach to countering health-related misinformation. These efforts highlight the need for an interdisciplinary approach in responding to misinformation.
Developing and implementing robust fact-checking mechanisms is critical to preserving the integrity of information and maintaining public trust in an increasingly connected world. Current methods for identifying and addressing misinformation are insufficient to handle its scale and complexity. Human fact-checkers struggle to process the vast volume of content generated on online platforms, and static fact-checking models often fail to adapt to the evolving tactics of misinformation spreaders. Additionally, the subjective nature of fact-checking introduces biases, particularly in cases involving framing or context manipulation. Detecting whether individuals are intentionally disseminating false information to incite others is essential for identifying the sources of misinformation and restoring public trust.
This project focuses on addressing the critical challenge of misinformation in the context of public health informatics. It aims to develop innovative analytical frameworks using LLMs. The first proposed approach targets the verification of truthfulness in health-related content. A generative artificial intelligence framework will be developed, leveraging LLMs with few-shot learning to construct semantic health knowledge graphs and enhance semantic reasoning. We will incorporate graph-based retrieval-augmented generation to address the hallucination issue common in LLMs and the limitations of static training data. By utilizing updated semantic health knowledge graphs, the framework will ensure that fact-checking is informed by the latest and most reliable medical news and health information. The second approach focuses on analyzing intentionally manipulated or fabricated content. Multi-agent LLMs will be employed to perform sentiment analysis, allowing for the identification of untrustworthy sources. This method will assess the intent behind misinformation and evaluate its potential impact on public health discourse. Our project seeks to provide robust solutions for automated fact-checking in the health domain. The outcomes of this research will contribute to public health informatics by improving the reliability of health information, mitigating the spread of false information, and fostering social trust in digital communication systems.
Project Reference No.: UGC/FDS23/M01/25
Project Title: The effectiveness of the 8-week GreenCare Programme in improving glycated haemoglobin levels, physical activity, and mental wellbeing in individuals with type 2 diabetes mellitus in Hong Kong: A randomised controlled trial
Principal Investigator: Dr KWOK Heather Hei-man (HKBU SCE)
Abstract
Background
Diabetes mellitus (DM) is a growing global health concern, with a prevalence projected to rise from 425 million cases in 2019 to 700 million by 2045. In Hong Kong, about 700,000 individuals have been diagnosed with diabetes, primarily type 2 DM (T2DM). Effective DM requires the control of biomarkers such as blood pressure, glycated haemoglobin (HbA1c), and body weight to prevent complications and delay disease progression. Although physical activity (PA) is vital for managing T2DM, physical inactivity is common, especially among older adults. In addition, consistent research has shown that a bidirectional relationship between chronic diseases and mental health conditions, with each influencing the other.
Objective
This proposed study will introduce the “GreenFit-DM Programme”, a nature-based, exercise focused intervention aimed at increasing PA levels and improving health outcomes among individuals with T2DM in urban settings. The primary aim of the proposed randomised controlled trial will be to evaluate the effectiveness of the GreenCare Programme over 8 weeks, focusing on its effects on HbA1c levels, PA, and mental wellbeing.
Study Design
The proposed study will be a randomised controlled trial involving 186 participants, aged 45 or above with T2DM at multiple community health centres in Hong Kong. Participants will be randomly assigned to either the intervention group or a wait-list control group. Data will be collected at five time points to assess changes in HbA1c levels, PA, and mental wellbeing.
Method
Based on the Health Belief Model, the “GreenFit-DM Programme” provides education, professional guidance, and customised exercise prescriptions to overcome barriers to PA. Participants will take part in structured sessions in accessible green spaces, supported by counselling and community-building initiatives. A pilot study conducted at Sham Shui Po District Health Centre demonstrated the feasibility of the programme, which yielded significant improvements in HbA1c levels and high participant satisfaction.
Significance
The aim of the proposed project will be to address a crucial research gap regarding T2DM management in Hong Kong and to offer insights into the effectiveness of nature-based interventions for promoting active lifestyles and improving health outcomes. The findings will enhance diabetes care and inform public health strategies for urban areas, ultimately reducing diabetes-related complications and improving the quality of life for individuals with T2DM.
Project Reference No.: UGC/FDS15/H09/25
Project Title: Sci-fi Cosmopolitanism: The Literary Reception of Western Science Fiction Novels in 1980s Hong Kong
Principal Investigator: Dr HO Ka-chun (Shue Yan)
Abstract
This project critically examines the reception of Western science fiction (sci-fi) in 1980s Hong Kong, revealing how it reflected and shaped local cultural identity. It shows that Hong Kong writers were drawn to sci-fi for its critique of imperialism and its ability to create “alternative realities”, which fostered critical dialogue about political and cultural dominance. This interest aligned with emerging ideas of cosmopolitanism. This project employs theories of aesthetic reception, cultural translation and magazine studies to achieve two objectives. First, by analyzing over 200 translations and essays from key Hong Kong literary magazines, the study demonstrates how Western sci-fi was actively integrated into Hong Kong’s literary field, especially how the influential authors such as Jules Verne, Isaac Asimov, and Philip K. Dick were reinterpreted within the aforementioned context by the translators and editors. Second, the research illustrates how Western sci-fi challenged traditional boundaries between lowbrow and highbrow literature, contributing to the broadening of Hong Kong writers’ cultural vision. Ultimately, the project offers a deeper understanding of how Western genres influenced local literary practices and contributed to the articulation of Hong Kong’s evolving cultural identity during the 1980s.
Project Reference No.: UGC/FDS14/E10/25
Project Title: A Multi-Agent Reinforcement Learning Framework for Optimizing E-Fulfillment Processes
Principal Investigator: Dr HO To-sum (HSUHK)
Abstract
E-commerce is experiencing substantial growth worldwide, yet most businesses in Hong Kong are small and medium enterprises (SMEs) and rely on rented storage spaces, making large investments in automation challenging. Artificial intelligence (AI) provides a practical alternative by optimizing resources through data-driven solutions. Currently, inefficient coordination among key e-fulfillment processes—stock receipt, replenishment, and order picking—often leads to delays and inflexibility, particularly when urgent orders emerge. Conventional optimization techniques are limited in adapting to these dynamic conditions, frequently requiring costly or ad-hoc adjustments.
This research proposes the use of a multi-agent reinforcement learning (MARL) model, which autonomously coordinates these fulfillment stages, identifies high-demand products, generates optimized picking routes, and allocates tasks to workers based on real-time priorities. The model’s flexibility allows it to be applied across various warehouse settings, making it particularly accessible for SMEs. By offering real-time, data-driven decision support, the proposed MARL model is positioned to enhance fulfillment efficiency and responsiveness in competitive logistics environments.
Project Reference No.: UGC/FDS14/E07/25
Project Title: Graph Neural Networks Assisted Dynamic Task Offloading in Internet of Vehicles
Principal Investigator: Dr HOU Yun (HSUHK)
Abstract
Project description:
The Internet of Vehicles (IoV) is transforming transportation by enabling vehicles to communicate with each other and with nearby roadside units (RSUs). This communication can enhance safety and efficiency on the road. However, because vehicles are constantly moving and their connections are always changing, deciding how to best share computing tasks among them is quite challenging. This project stands out by exploring the unique potential of sharing computational power among vehicles, rather than solely offloading tasks to roadside infrastructures. This approach makes decision-making even more challenging, as the contact window between two vehicles is often unpredictable due to their high mobility. Our main goals include creating a system that can predict how long two vehicles will stay connected. This information is crucial for helping vehicles decide when and where to offload tasks. We will also use Graph Neural Networks (GNNs) to analyse the network of vehicles, which will help us understand how to best allocate communication resources and timings for tasks that need to be shared. The insights gained from these analyses will improve the decision-making process for the Deep Reinforcement Learning (DRL) based task offloading schemes.
The project is organized into three key objectives. First, we will develop a prediction framework for estimating how long vehicles will be in contact with RSUs and other vehicles, which is essential for making informed offloading decisions. Second, we will create a GNN model to learn the relationships between network entities and efficiently manages communication resources between them. Finally, we will build a flexible DRL framework that can adapt to the ever-changing network conditions, allowing vehicles to share tasks not only with RSUs but also with each other.
The project team:
The background and research experience of the Principal Investigator (PI) and Co-Investigators (Co-Is) put them in a good position to carry out the proposed studies. The PI has solid research experience, both theoretically and practically, in Wireless Network Resource Allocation and Vehicular Networks. The Co-Is with ample experience in Wireless Edge Computing and Artificial Intelligence, especially on GNN and RL, compensate for the PI’s expertise gap. In terms of project management experience, the PI coordinated one Innovation and Technology Fund (ITF) project and one FDS project on V2X communications, which delivered commercial-grade reference designs of LTE-V2X systems and distributed vehicle co-movement algorithms respectively. The PI is also managing an FDS project aimed at enhancing model freshness with extreme cases in autonomous driving. Such project management experience and the research strengths of the team serve as a solid foundation for the successful delivery of the proposed project.
Significance of the Project:
If successful, this research project will establish a foundational framework for future computational task offloading in vehicular networks, shifting the paradigm towards sharing tasks among surrounding vehicles rather than solely relying on roadside units. This will expand the range of task offloading options and enhance decision-making processes, ultimately enabling vehicles to efficiently and timely execute computational tasks that are critical for safe navigation.
Project Reference No.: UGC/FDS13/E03/25
Project Title: Acoustic-based automated obstructive sleep apnea detection method
Principal Investigator: Dr HSUNG Tai-chiu (Chu Hai)
Abstract
Obstructive sleep apnea (OSA) is a prevalent sleep disorder marked by the collapse of the upper airway during sleep. While it predominantly affects adult men with obesity, it can impact individuals of all ages and body types. A 2019 study estimated that 936 million adults worldwide suffer from mild to severe OSA, with China experiencing the highest prevalence at 745 million. Most cases remain undiagnosed, contributing to daytime sleepiness, snoring, and heightened health risks.
Polysomnography (PSG) is the gold standard for diagnosing OSA; however, its accessibility is constrained by discomfort and the need for professional assistance. This underscores the necessity for a more convenient and affordable diagnostic method.
Our recent study analyzed 66 PSG and voice data sets, revealing that patients with OSA display unique high-frequency sound features in their voices compared to those without the disorder. These features enhanced the accuracy of traditional speech-based OSA detection from 80.3% to 84.85%.
To enhance the accessibility of OSA detection for the community, we aim to advance our research efforts. Specifically, we plan to create a dataset of 600 PSG tests paired with high-fidelity sound recordings, identify optimal acoustic features for automatic detection through digital signal processing and AI, and develop mobile applications that facilitate easy acoustics-based OSA detection.
Project Reference No.: UGC/FDS16/M31/25
Project Title: Intelligent and Reconfigurable Rehabilitative Device Based on Multi-modal Fusion and Advanced Neural Network
Principal Investigator: Dr HUANG Hongli (HKMU)
Abstract
The growing demand for advanced assistive technologies, particularly in prosthetics and rehabilitation, underscores the need for intelligent systems that can accurately recognize human motion intent in real-world environments. Prosthetic limbs, exoskeletons, and rehabilitation devices that seamlessly respond to the user’s intended movement are essential for improving the quality of life for individuals with physical disabilities. However, existing systems still struggle with issues related to robustness, adaptability, and long-term usability, often requiring complex recalibration when used across different limb positions, physical conditions, and dynamic environments. This proposal seeks to address these challenges by developing next-generation intelligent assistive devices based on multimodal sensor fusion and advanced neural networks, significantly enhancing the performance, personalization, and adaptability of prosthetics and rehabilitation technologies.
The key deliverables of this project are focused on creating a robust and intelligent platform for prosthetic devices, aimed at improving their functionality, responsiveness, and ease of use. In the preliminary work, we’ve developed a state-of-the-art multimodal fusion framework for motion intent recognition that integrates electromyography (EMG) and force myography (FMG) signals using advanced sensor fusion techniques (Huang et al, IEEE JBHI, under review). This framework has been proved to be highly effective to enhance the long-term repeatability of the motion intent recognition system across different limb positions, maintaining over 90% accuracy in the cross-day test after a week without any recalibration. Building on these encouraging preliminary results, we are poised to extend our work into a fully integrated system that not only recognizes user intent with high precision but also adapts dynamically to real-world variability. In the next phase of our research, we will leverage these foundational innovations to develop a suite of deliverables that further enhance personalization and adaptability.
By integrating our proven techniques with cutting-edge meta-learning and sensor fusion strategies, we aim to construct an intelligent motion intent detection module and a Complementary Pyramid Fusion Siamese Network (CPFSN) capable of rapid user-specific adaptation. The main deliverables of the proposed project are: (1) A Complementary Pyramid Fusion Siamese Network (CPFSN) designed to enable rapid personalization of assistive devices. This meta-learning model will allow assistive devices to adapt to new users and motion patterns with minimal data, significantly reducing the need for retraining and making the devices more user-friendly and adaptable. (2) An intelligent motion intent detection module that can be integrated into multiple assistive devices, such as prosthetics and exoskeletons. This module will allow these devices to recognize and respond to user movements with high accuracy, enhancing their functionality and making them more versatile across different applications. (3) A reconfigurable, lightweight, and low-cost 3D-printed prosthetic limb that incorporates the multimodal fusion framework, CPFSN model and the intelligent motion intent detection module. This prosthetic will be fabricated using additive manufacturing, ensuring affordability and customization, and will undergo real-world trials to evaluate its performance, durability, and user satisfaction. The impact of this project extends beyond technological advancements in prosthetics and rehabilitation.
According to the World Health Organization (WHO), over 30 million people worldwide require prosthetic or orthotic devices, a number projected to rise in response to population aging, escalating rates of diabetes and vascular diseases, and other contributing factors. By improving the robustness and adaptability of assistive devices, this research has the potential to transform the lives of individuals with disabilities, offering them greater independence and enhancing their physical rehabilitation outcomes. The combination of multimodal sensor fusion, neural networks, and meta-learning offers a new paradigm in the design of assistive devices, setting a foundation for future innovations in the field.
Project Reference No.: UGC/FDS24/B14/25
Project Title: The Role of Social Media in Unveiling Corporate Greenwashing: Evidence from Hong Kong
Principal Investigator: Dr HUANG Wenli (PolyU SPEED)
Abstract
With a growing emphasis on environmental sustainability in the global market, the issue of greenwashing has attracted significant attention from both academics and industry. The term “Greenwashing” refers to the practice of portraying business as environmentally friendly either by providing misleading information or giving a false impression on the sustainability of a product or service (e.g., Delmas & Burbano 2000; Kim & Lyon 2015; Montgomery, Lyon, and Barg 2023). A global review led by International Consumer Protection and Enforcement Network (2021) found that 40% of green claims made online could be misleading. Prior literature has documented various drivers of greenwashing, including regulatory pressure, market-level competition, and firm characteristics such as growth opportunities and board compositions. The literature also examines the consequences of greenwashing. Despite limited evidence, greenwashing is generally viewed negatively by the public, with academic commentators raising concerns about the authenticity of net zero commitments (Guinson, 2021).
The rapid growth of social media has created powerful and dynamic communications between firms and their stakeholders. While literature has examined the role of social media in corporate disclosure in general, little research has been done on the role of social media in shaping individuals’ perceptions about firms’ ESG practices. The proposed study aims to investigate the relationship between social media and greenwashing using a sample of firms listed in Hong Kong. Hong Kong offers a unique setting for this research due to its mandated ESG reporting since 2015. The widespread use of social media as voluntary disclosure channel presents a tension worth examining: does social media platform encourage or deter corporate greenwashing under a mandatory reporting regime?
The rapid growth of social media has created powerful and dynamic communications between firms and their stakeholders. While literature has examined the role of social media in corporate disclosure in general, little research has been done on the role of social media in shaping individuals’ perceptions about firms’ ESG practices. The proposed study aims to investigate the relationship between social media and greenwashing using a sample of firms listed in Hong Kong. Hong Kong offers a unique setting for this research due to its mandated ESG reporting since 2015. The widespread use of social media as voluntary disclosure channel presents a tension worth examining: does social media platform encourage or deter corporate greenwashing under a mandatory reporting regime?
Moreover, this study will have implications for ESG regulations and practice. As greenwashing becomes more prevalent globally, many countries and regions are seeking measures to tighten the scrutiny and transparency of ESG disclosure. Hong Kong is no exception. The findings of this study will provide valuable insights into the public awareness of corporate greenwashing and inform the policy makers of the potential mechanism to combat it, aligning with the Hong Kong Special Administration Region (HKSAR)’s strategic goal of promoting Hong Kong as a green and sustainable finance hub.
Project Reference No.: UGC/FDS16/E23/25
Project Title: Prognosis of Alzheimer’s Disease in Early Mild Cognitive Impairment Patients Using Quaternion-based Multi-Frequency Dynamic Functional Connectivity Analysis of Resting-state fMRI
Principal Investigator: Dr HUNG Kevin King-fai (HKMU)
Abstract
The global acceleration of population ageing has led to a significant rise in chronic diseases, including dementia, which affects approximately 55 million people worldwide and incurs an annual cost of USD 1.3 trillion. In Hong Kong, it is estimated that over 100,000 people are living with dementia. Alzheimer's disease (AD), the most common form of dementia, poses a substantial burden on healthcare systems due to its progressive nature and lack of definitive cure. Early detection of AD is crucial for timely interventions that can slow disease progression, improve patient outcomes, and reduce long-term healthcare costs. However, current diagnostic methods, particularly in the early stages of the disease, are often inaccurate, with misdiagnosis rates as high as 70% in primary care settings, as reported by the Alzheimer’s Association in 2021. Mild Cognitive Impairment (MCI) is a precursor to AD, with one-third of MCI patients progressing to AD within five years. However, some people with MCI do not experience further cognitive decline and may even revert to normal cognitive function, making the progression from MCI to AD unpredictable. Accurate prediction of which MCI patients will develop AD is essential for improving life quality and reducing healthcare costs. Recent advancements in machine learning have shown promise in enhancing diagnostic accuracy, but these methods typically rely on clinical data collected after the onset of dementia symptoms, limiting their effectiveness for early detection.
Functional connectivity (FC) analysis based on resting-state functional magnetic resonance imaging (rs-fMRI) data provides a promising solution for early AD detection. FC examines the relationships between neural activities in different brain regions and can identify changes before clinical symptoms appear. However, traditional FC analysis faces several challenges that hinder its overall performance. Most fMRI studies on brain diseases have focused on a specific frequency band or combined multiple frequency bands through simple concatenation or linear combination. These methods do not fully capture the relationships between different frequency bands, reducing their effectiveness in prognosis. Additionally, traditional FC analysis typically focuses on correlations between any two brain regions, which degrades the resulting brain connectivity due to the loss of multiple-region relationships. Furthermore, most fMRI studies assume that FC remains static, overlooking the temporal variability of brain connectivity during MRI scans. FC-based features often cause overfitting problems in machine learning, and their redundant information hampers classification results. These bottlenecks challenge more accurate prediction from MCI progression to AD in clinical settings.
Quaternion signal processing is a promising technique that has the potential to tackle these problems. Quaternion signal representation preserves the interrelationship among different frequency bands, and quaternion manipulation enables the extraction of inherent features for better FC analysis. To the best of our knowledge, work in this proposed project is the first time a quaternion-valued representation of rs-fMRI signals with multiple frequency bands is introduced. Additionally, no research has yet explored using Quaternion Least Absolute Shrinkage and Selection Operator (QLASSO), Quaternion Principal Component Analysis (QPCA) or Quaternion Linear Discriminant Analysis (QLDA) for the manipulation of dynamic FC analysis of brain networks. Considering these knowledge gaps, the objectives of this project are to i.) develop a novel quaternion-valued representation that integrates multiple frequency bands of rs-fMRI signal and preserves their interrelationships; ii.) perform QLASSO to construct a fully connected dynamic FC with a sliding-window approach for understanding inherent brain states in MCI prognosis; iii.) execute QPCA and QLDA for feature extraction and selection for the prediction of MCI progression to AD; and iv.) demonstrate the classification of MCI-converted and MCI-stable conditions based on the new features. Achieving these objectives will provide a more accurate and efficient technique for MCI prognosis, ultimately contributing to early detection of AD.
Project Reference No.: UGC/FDS16/B04/25
Project Title: The Voice Catchers: How Female Entrepreneurs Leverage Vocal Features to Shape Funder Decision-Making
Principal Investigator: Dr JI Li (HKMU)
Abstract
Female entrepreneurs face significant challenges in securing funding due to inherent gender biases. Traditional funding mechanisms often overlook the nuanced ways in which female entrepreneurs can strategically present themselves to potential investors. While machine learning technologies have enabled researchers to explore the relationship between vocal features and audience perception, there is a lack of systematic studies on how female entrepreneurs can leverage these vocal features to overcome gender bias and enhance their chances of securing funding.
To fill these research gaps and offer practical solutions, this proposal includes three studies: In Study One, we will identify vocal features that are perceived as gender stereotypes for female versus male entrepreneurs through online surveys of experienced funders. In Study Two, we will develop hypotheses on the effect of female entrepreneurs’ vocal features on funder perception and decisions, then empirically test these hypotheses based on lab experiments. In Study Three, we will use field data from Kickstarter to validate the relationship between female entrepreneurs’ vocal features and funding outcomes.
This study on the strategic use of vocal features by female entrepreneurs will contribute significantly to the fields of entrepreneurship, communication, and gender studies. On one hand, it will bridge the gap between emerging literature on vocal features and traditional entrepreneurial funding research by identifying and validating key vocal features that influence audience perception and funding success, particularly for female entrepreneurs. On the other, our study will extend communication literature by empirically examining the mechanisms through which vocal features shape funder perception and funding decisions.
Managerially, the outcomes of this study will have important implications for various stakeholders in the entrepreneurial ecosystem. For female entrepreneurs, this study will provide actionable insights on how to strategically use vocal features to enhance their pitch delivery and improve their chances of securing funding. For investors and funding organizations, this study will underscore the importance of considering dynamic vocal features in their evaluation processes. For policymakers and government bodies, the findings will offer valuable insights into developing policies and initiatives aimed at supporting female entrepreneurs.
Overall, this study aims to create new knowledge that not only advances academic research but also provides practical strategies for female entrepreneurs, investors, and policymakers to foster a more inclusive and equitable entrepreneurial landscape for Hong Kong and the world.
Project Reference No.: UGC/FDS24/E20/25
Project Title: Developing Knowledge Driven Multimodal Framework for 3D MRI Analysis and Interpretation: Grounding Visual Language Models for Diagnostic Reasoning
Principal Investigator: Dr KHAN Sheheryar (PolyU SPEED)
Abstract
3D Magnetic Resonance Imaging (MRI) plays a key role in disease diagnosis, treatment planning, and monitoring. Advances in deep learning (DL) techniques have shown significant potential in automating MRI-based diagnoses. However, the interpretability and generalization capabilities of DL models in the MRI domain remain challenging. Convolutional neural networks (CNN) based DL models are often black boxes, lacking reasoning and difficult to interpret. Clinicians and radiologists need interpretability to trust and collaborate effectively with AI systems. While dealing with generalization, one major issue is domain shift, which occurs when deploying DL models across different clinical settings or patient populations. Models trained on one dataset often degrade on another due to variations in imaging protocols, pulse sequences, equipment, demographics, and disease characteristics.
Vision language models offer a suitable solution for interpretation. However, general-purpose Vision Language Models (VLMs) struggle with reasoning with multi-sequence MRI data due to their inability to extract crucial small details spanning 3D slices in MRI. We argue that effective analysis of such data requires localized cues to capture better anatomical and clinical details called grounding, which can be modelled through robust knowledge distillation segmentation methods coupled with explicitly prepared radiomics. For instance, 3D-knee osteoarthritis (OA) MRIs comprise thin tissues such as cartilage and meniscus, which require advanced insights. Radiologists rely on cartilage continuity across slices, joint spacing, or synthesizing tissue surface area to diagnose OA severity. Current VLMs lack the capability to process these radiomics-based features or render 3D thickness maps, which impacts the solid reasoning behind derived answers, therefore limiting their clinical applicability.
This proposal presents a generalized multimodal framework to address the domain variations and interpretation in MRI understanding. We primarily focus on knee osteoarthritis (OA) MRI analysis and interpretation while also demonstrating the applicability of our approach to liver MRI analysis, showcasing its ability to handle varied anatomical structures. Our framework is built up of three key units: Knowledge Distillation unit (KDU), Grounding unit (GU), and Vision Language Understanding and Reasoning unit (VLURU).
KDU presents a novel semi-supervised segmentation approach using Successive Eigen Noise-assisted Mean Teacher Knowledge Distillation SEN-MTKD. This method addresses domain shifts and label shortages while improving segmentation accuracy for thin tissues like cartilage and meniscus. Meanwhile, unsupervised approaches to carry out the attention maps in MRI images are presented to facilitate region detection.
GU is designed to provide strong grounding to VLM. We propose anatomical and radiomics-aware alignment in VLMs through grounding, which involves precise segmentation labels for key anatomical structures, such as the liver in abdominal imaging and knee joint tissues, including the meniscus and cartilages. These segmentation outputs enable extracting clinically relevant radiomics features, such as surface areas, tissue volumes, inter-tissue gaps, and a detailed set of metrics that characterize tissue geometry and morphology. Furthermore, 2D thickness maps derived from 3D segmented tissue offer a richer spatial representation and visual clue of tissue structures.
VLURU comprises a two-step VLM, combining structured knowledge graphs and domain-specific logical reasoning. The causal reasoning module, initially trained on large-scale medical datasets, generates knowledge graphs from visual observations and integrates them with logical rules to produce accurate, explainable answers to medical queries. VLURU uses GU-derived cues to fine-tune VLM, enabling it to generate clinically relevant and contextually accurate responses to queries about MRI images. This multi-modal approach bridges the gap between pixel-level imaging data and high-level language-driven reasoning, creating a coherent framework for automated medical image interpretation.
Project Reference No.: UGC/FDS16/B09/25
Project Title: Responding to Radical Innovation: Strategic Approaches to Knowledge Prioritization in Outsourcing
Principal Investigator: Dr KHURSHID Faisal (HKMU)
Abstract
In the competitive and ever-evolving landscape of business, knowledge stands as a crucial strategic asset for achieving performance advantages. When radical innovations emerge, incumbent firms may lack the expertise to address new challenges. Such technological changes can disrupt markets, rendering existing knowledge and capabilities obsolete, as seen with Nokia, Motorola, and BlackBerry in the early smartphone era. To adapt, firms can (i) develop new products in-house after conducting thorough research and development, (ii) acquire knowledgeable suppliers, or (iii) form outsourcing relationships. The first two options are time-consuming and require significant resources, making outsourcing a viable survival strategy. Some scholars suggest that exclusively leveraging suppliers' knowledge related to outsourced components can help firms cope with challenges of radical technological changes. However, other scholars highlight the risks of over-relying on suppliers' knowledge and emphasize the benefits of retaining in-house knowledge for outsourced components to control suppliers’ opportunism and better understand underlying technological changes. In practice, firms can use both internal and external (supplier) knowledge sources. They can prioritize one source over the other based on market conditions and technological challenges and re-prioritize as technological evolution progresses. Nevertheless, research on the relative importance of the firm's knowledge stock and its suppliers' knowledge stock over different phases of technological change remains scarce.
This research examines two phases of technological change: the pre-dominant technology phase, where multiple technologies coexist and create uncertainty about the best features and materials, and the post-dominant technology phase, where a dominant technology emerges, reducing uncertainty and emphasizing firm-specific design features. This study aims to investigate how a firm's and its suppliers' knowledge affect product performance during both phases of radical innovation and identify which knowledge source is more beneficial in each phase. Additionally, firms can decide which type of knowledge to retain in-house and focus their internal research and development efforts on either architectural knowledge or component knowledge. Architectural knowledge involves understanding how different components of a system interact and integrate, providing a holistic view of the system's design. In contrast, component knowledge focuses on specific individual parts of the system. Considering strategic needs, firms may prioritize one type of knowledge over the other. The research will also examine how retaining in-house architectural vs. component knowledge impacts performance.
Drawing upon the concepts of hierarchy hazards and exchange hazards, which are derived from the behavioral theory of the firm and transaction cost economics, this study examines how these hazards distinctively impact firms during technological change. Hierarchy hazards arise from established routines that favor the status quo, leading to organizational inertia and limiting innovation. Exchange hazards involve opportunism in supplier relationships, increasing transaction costs. The different technological and transactional demands in the pre- and post-dominant technology phases expose firms to different risks. Managing knowledge related to outsourced components, whether in-house or externally controlled, significantly impacts a firm's ability to mitigate these risks. The study will investigate a high-tech industry using data at both firm and product levels. A multilevel modeling approach will be employed to analyze the data. This study aims to predict that prioritizing supplier knowledge in the pre-dominant phase and in-house knowledge in the post-dominant phase could offer performance advantages in response to radical innovations. Additionally, retaining in-house architectural knowledge can provide strategic benefits compared to component knowledge.
Project Reference No.: UGC/FDS16/E04/25
Project Title: A Novel Distributed Computing System with Gradient Coding Schemes for Machine Learning
Principal Investigator: Dr KWAN Ho-yuet (HKMU)
Abstract
Artificial Intelligence (AI) revolutionizes every aspect of our lives. The essence of AI is its power on inference and prediction. Such power is attributed to one of the various technological advancements in the past few decades - Machine Learning. An accurate machine learning model is needed for unleashing the power of AI. However, it involves massive data samples and computation resources to train a machine learning model. Recently, a paradigm known as distributed machine learning has emerged, where a master computing node divides a large computation task into smaller ones and distributes them to edge devices within a network. This approach allows for scalable computation power in a flexible and manageable manner. However, there are some issues that need to be addressed for its full implementation.
One of the prominent challenges in implementing distributed machine learning is the heterogeneity of the computation power of each edge device that results in the presence of slow edge devices called stragglers. Since a master node basically must wait for results returned from all edge devices before proceeding to the next round computation, stragglers would severely limit the efficiency of a machine learning process. Another issue of distributed machine learning is security. Results computed by edge devices are required to be sent back to the master node for combining. The resultant machine learning model might be ruined if some edge devices intentionally send malicious results to the master node.
Gradient coding is an effective solution to both challenges. By distributing edge devices overlapping training data blocks from the whole training dataset, techniques from coding theory which introduces data redundancy among edge devices can be used to complete an original machine learning process even there are some unresponsive and/or malicious edge devices in the network. However, existing distributed machine learning systems with gradient codes put emphasis mainly on optimizing the completion time of training. From a system design perspective, a trade-off between training completion time and accuracy is desirable to enhance implementation flexibility. In addition, existing measures that are resilient against adversarial-attacks are not integrated well into a distributed machine learning system with gradient codes. That may undermine the effectiveness of gradient codes.
In this proposed project, our team will design a low implementation complexity distributed computing system with gradient coding schemes for machine learning to optimize computation time and provide a trade-off between the accuracy of a gradient computation and its computation time. We will also design a training data block re-allocation scheme to handle the cases of the heterogeneity of computation rate of worker nodes so that computation time can be further optimized. Finally, we will develop an adversarial-attack-resilient mechanism and fully integrate it into our proposed system for the protection of a distributed machine learning process. We anticipate that our design, characterized by high performance, low implementation complexity, and adversarial robustness, will serve as a reference for future standardization of distributed machine learning in the industry.
Project Reference No.: UGC/FDS16/P08/25
Project Title: A Novel Vision Transformer-Based Neural Network for Image Classification and Jet Tagging in Lepton-Flavor Violating Processes
Principal Investigator: Dr KWOK King-wai (HKMU)
Abstract
This research project addresses a critical and urgent challenge in high-energy physics (HEP) experiments: efficiently utilizing the ever-increasing volume of collected data, both present and future. The difficulty lies in analyzing this data within a reasonable timeframe and with acceptable uncertainty, requiring careful compromises to find optimal solutions. Specifically, the ATLAS experiment at the Large Hadron Collider (LHC) collected proton-proton collision data with an integrated luminosity of 140 fb-1 in Run 2 (2015-2018). This is expected to double to 300 fb-1 in Run 3 (2022-2025), further exacerbating the data analysis challenge.
Using machine learning (ML) to analyze high-energy physics (HEP) experimental data is an established concept, yet its full potential remains untapped. Interest in applying ML to HEP surged towards the end of Run 2, leading to publications such as a white paper (Kim et al., 2017) and various review articles (Guest, Cranmer & Whiteson, 2018; Kheddar et al., 2024). A Kaggle competition (Strong, 2020) even engaged the public in exploring ML for HEP analysis.
Convolutional Neural Networks (CNNs) utilize hierarchical feature extraction, progressing from local patterns to complex representations. A study on hadron calorimeter energy deposition patterns for prompt and displaced jets demonstrated CNNs' superior performance over Boosted Decision Trees (BDTs) in background rejection. At 60% signal efficiency, CNNs achieved a 93% background rejection rate compared to 69% for BDTs (Bhattacherjee, Mukherjeeb & Senguptaa, 2019). Another study employing a CNN-based Deep Neural Network (DNN) classified five jet types (top quark, W boson, Z boson, gluon, and light quark) with over 89% accuracy for the first three and over 60% for the latter two, exceeding BDT performance (Sandoval, Manian & Malik, 2024).
Originally developed for natural language processing, the Transformer model (Vaswani, 2017), with its multi-head attention mechanism, was adapted for computer vision in 2019 as the Vision Transformer (ViT) (Ramachandran et al., 2019). ViTs have demonstrated competitive, and often superior, image classification performance compared to traditional CNNs, particularly with sufficient training data. One study showed that fine-tuned ViTs achieved four times higher top-1 accuracy than CNNs (Paul & Chen, 2021). While possessing higher capacity, ViTs can be less data-efficient than CNNs. However, subsequent variants like DeiT (Touvron et al., 2021) and PVT (Wang et al., 2021) have addressed these limitations, improving efficiency, accuracy, and domain-specific applicability. The recent development of a massive 113-billion parameter ViT for weather and climate prediction (Wang et al., 2024) further highlights the potential of ViTs. Given the large datasets characteristic of HEP experiments, applying ViTs and their variants to tasks like jet analysis is a promising avenue for performance gains. This potential is underscored by a recent study demonstrating a 1.4% accuracy improvement (to 85.2%) using the ViT-based ParT architecture for jet tagging (Qu, Li & Qian, 2024).
This research aims to develop Vision Transformer (ViT) architectures for both image classification and jet tagging in lepton-flavor violating (LFV) processes. These processes, allowed in beyond the Standard Model (BSM) physics, are predicted by models such as R-parity violating Supersymmetry (RPV SUSY), Z' boson model, and quantum black holes (QBH) models. Performance will be compared against CNNs, as used by Kim et al. (2023), and traditional Boosted Decision Tree (BDT) based cut-based methods (Kwok, 2021).
Kim et al. (2023) used a CNN-based image classification architecture to study RPV SUSY events, achieving a 1.845-fold increase in signal efficiency and a 1.2 times increase in expected significance compared to cut-based methods. Given that ViTs generally outperform CNNs on large datasets, and CNNs typically outperform BDT-based analyses, it is conservatively estimated that using a ViT-based architecture for image classification in LFV studies could yield at least a three-fold improvement in signal efficiency and a two-fold improvement in expected significance over traditional cut-based methods. Furthermore, employing a ViT-based architecture for jet tagging is expected to enhance performance even further.
Project Reference No.: UGC/FDS16/M25/25
Project Title: Do anti-amyloid beta monoclonal antibodies reassuringly offer hope for Alzheimer’s disease risk reduction?
Principal Investigator: Dr KWOK Maggie Man-ki (HKMU)
Abstract
Dementia primarily Alzheimer’s disease (AD) poses challenges to global healthy aging. In June 2021, Aducanumab was the first-ever drug for AD treatment approved by the United States Food and Drug Administration (U.S. FDA). Controversies over this long-awaited AD drug remain including the questionable clinical benefit of delaying cognitive and functional decline, and any unknown longer-term adverse effects beyond amyloid-related imaging abnormalities. A post-marketing trial was required by the U.S. FDA to verify the clinical benefits of aducanumab, but was terminated by the pharmaceutical company. Nonetheless, Lecanemab and Donanemab, two other medications in the same drug class as aducanumab, received FDA approval in 2024.
Genetic validation using Mendelian randomization (MR) will provide evidence that is urgently needed to clarify the potential effects of Aducanumab, Lecanemab and Donanemab so as to inform clinical decisions and strategies for preventing dementia. This study will clarify the role of anti-amyloid beta monoclonal antibodies including Aducanumab, Lecanemab and Donanemab in AD using MR and identify its possible longer-term side effects using a phenome-wide Mendelian Randomization study (MR-PheWAS).
Objectives: To assess the role of anti-amyloid beta monoclonal antibodies including Aducanumab, Lecanemab and Donanemab in AD using a two-sample MR study and to discern their pharmacovigilance using an MR-PheWAS.
Design: Two-sample MR study and MR-PheWAS.
Participants: For the two-sample MR study, genome-wide association studies (GWAS) of amyloid PET imaging (n=11,556 people of European descent and n=1,494 people of primarily East Asian descent), GWAS of AD (n=455,258) and cognitive function (n=23,066) in Westerners, and GWAS of dementia in East Asians (Japanese) (n=8,036) will be used. For the MR-PheWAS, the UK Biobank (n=~500,000) and BioBank Japan (n=~200,000) will be used.
Exposure: Genetic mimics for anti-amyloid beta monoclonal antibodies including Aducanumab, Lecanemab and Donanemab based on genetic variants mimicking their effects (i.e., amyloid protein reduction) as instruments.
Outcomes: AD and cognitive function as the primary outcomes for efficacy; a wide range of diseases and related traits as the secondary outcome for safety.
Data analysis: Estimates for genetically mimicked anti-amyloid beta monoclonal antibodies on AD will be based on inverse variance weighting estimates with multiplicative random effects, with sensitivity analyses including weighted median, MR-Egger, MR-ConMix, MR-RAP, MR-PRESSO, and multivariable MR. Any associations of genetic mimics for anti-amyloid beta monoclonal antibodies with other diseases and traits will be identified from the MR-PheWAS. Colocalization will also be applied.
Expected results: Genetically mimicked anti-amyloid beta monoclonal antibodies including Aducanumab, Lecanemab and Donanemab may not be associated with AD in Westerners or East Asians. They might have more side effects in Westerners than East Asians.
Significance: These findings will inform potential clinical efficacy and pharmacovigilance of anti-amyloid beta monoclonal antibodies for benefit-risk assessment, thereby generating timely evidence to assist clinical decisions on the use of Aducanumab, Lecanemab and Donanemab in older adults at risk of AD.
Project Reference No.: UGC/FDS16/H42/25
Project Title: Automatic Generation of Scenario-based E-learning Games for Nursing Students: A Novel Approach Using GenAI Agents
Principal Investigator: Dr KWOK Tai-on (HKMU)
Abstract
Training nursing students using e-learning games with simulated scenarios is an important alternative to conventional teaching methods due to its high accessibility (students can learn anytime and anywhere at their own pace) and modalities (interactive learning with rich media contents). But creating such learning simulations will typically take considerable amount of faculty time in the design and development process, and incur a development cost. In this project, we propose to develop and evaluate using generative artificial intelligence (GenAI) for automating the making of scenario-based e-learning games for nursing students. Specifically, we will develop two key innovations using GenAI:
1) GenAI agents (or assistants) that interview nursing teachers (or related course team members) in plain English to understand the details of the patient case they want to simulate via a web application that we shall build. The GenAI agents will turn these conversations into a complete storyline for the game scenario.
2) GenAI-based toolchains that can automatically build the game software from the storyline. The GenAI agents will generate all the 2D scenes, character animations, sounds, and programming code needed for an interactive game, which will be an HTML5-based web game to be run in a web browser on a PC, laptop, tablet or mobile phone. The scenario games will be reviewed by the teachers and improved before trial use by students.
Our design tools using GenAI to do most of the time-consuming manual work involved in game development would free up teachers to focus on instructional design. This approach can also make creating multiplayer simulated patient games much faster and cheaper. In this project, we will also develop a learning management system (LMS) to provide personalized learning and assessment for students using the GenAI-generated games. On the other hand, in this project we will harness the power of GenAI for creation of scenario-based e-learning games for the 12 areas of Principles and Practice of Nursing outlined by The Nursing Council of Hong Kong.
We will use a combination of the constructivist learning theory, situated learning theory and motivation and engagement theories as a guiding framework for the development and evaluation of our proposed GenAI framework: (1) Constructivism: learners construct their own understanding and knowledge of the world through experiences and reflecting on those experiences. Scenario-based e-learning games provide interactive and immersive experiences that allow nursing students to engage in realistic scenarios. This helps them build knowledge by actively participating in problem-solving and decision-making processes. (2) Situated learning: learning is most effective when it is situated in the context in which it is applied. Scenario-based games place students in authentic clinical environments, helping them understand the context and apply their knowledge in practical settings. (3) Motivation and engagement: theories such as Self-Determination Theory emphasize the importance of intrinsic motivation and engagement in learning. Games are inherently engaging and can motivate students through elements like challenges, rewards, and storytelling. This increased motivation can lead to better learning outcomes. We will conduct a comprehensive evaluation of the GenAI framework in the following aspects: usability, efficiency, and educational outcomes.
If successful, this project will make scenario-based e-learning widely accessible for nurse training. Our GenAI techniques could also be adapted to create e-learning games for medical, physiotherapy and other healthcare students, to help train healthcare workers to provide quality patient care. In addition, the completion of the project will bring broader scholarly contributions to other fields such as engineering, education, and computer science.
Project Reference No.: UGC/FDS16/M04/25
Project Title: Unravelling the Role of miRNAs in Transgenerational Male Reproductive Impairment in Marine Medaka Fish Induced by BDE-47
Principal Investigator: Dr LAI Keng-po (HKMU)
Abstract
Endocrine-disrupting chemicals (EDCs) can disturb endocrine systems of animals, leading to reproductive impairment and malformation and pose a significant threat to the sustainability of various species, despite occurring in very low concentrations (ppt) in the environment. Research has shown that some EDCs can affect not only the exposed individuals but also cause transgenerational adverse effects, even in offspring that have never encountered these chemicals. This group of EDCs may modify the epigenome or miRNA, leading to lasting impacts across generations. Polybrominated diphenyl ethers (PBDEs), commonly used as flame retardants, have become widespread in the global environment. PBDEs may adversely affect the reproduction and development of fish and higher vertebrates, including humans. miRNAs, which control gene expression transcriptionally, are recognized as epigenetic modulators. Our preliminary transgenerational study revealed that parental (F0) marine medaka fish exposed to 2,2’,4,4’-Tetra-bromodiphenyl ether (BDE-47), a common PBDE congener, could impair sperm motility of subsequent generations (F1-F2), despite these progenitors have never been exposed to BDE-47. Small RNA sequencing analysis of sperms from each generation (F0-F2) indicated common deregulation of a cluster of miRNAs reported to play roles in sperm motility, including the upregulation of miR-28-3p, miR-182, and miR-24-3p, alongside the downregulation of miR-296 and miR-200. These dysregulated miRNAs may play a crucial role in transgenerational male reproductive impairment induced by BDE-47. This proposal aims to test the hypothesis that parental exposure (F0) to BDE-47 alters a cluster of miRNAs targeting the genes responsible for maintaining sperm motility, resulting in the transgenerational reproductive impairment in subsequent generations (F1 to F2) in male fish. To test this hypothesis, we propose to use the in vivo medaka fish (Oryzias melastigma) and in vitro medaka fish testicular cells coculture as models to: (1) identify the gene targets of miR-28-3p, miR-182, miR-24-3p, miR-296, and miR-200 responsible for maintaining sperm motility; (2) determine the direct binding of miR-28-3p, miR-182, miR-24-3p, miR-296, and miR-200 to the identified gene targets related to sperm motility; and (3) reveal the role of miRNA-mRNA pairs in transgenerational impairment of sperm motility induced by BDE-47. The findings of this proposal will unravel the underlying mechanisms of transgenerational male reproductive impairment induced by PBDE exposure, a common pollutant in the global aquatic environment. By exploring how miRNAs changes contribute to sperm quality across generations, this study offers valuable insights into the full scope of pollutant-induced transgenerational reproductive impairments. These findings have implications for policy-making, risk assessment, and intervention strategies designed to mitigate the adverse effects of pollution on reproductive outcomes. Given that regulation of both miRNAs and reproductive functions are highly conserved in vertebrates, the results of this study may also shed light on higher vertebrates, including humans.
Project Reference No.: UGC/FDS11/B04/25
Project Title: Agricultural tourism for rural resilience: A sustainable livelihoods approach
Principal Investigator: Dr LAI Tin-hang (SFU)
Abstract
This study aims to develop a framework for agricultural tourism to achieve sustainable rural resilience that incorporates the principles of Sustainable Livelihoods Approach (SLA) alongside a two-dimensional assessment approach focused on livelihood diversity and livelihood freedom. To achieve this, the research employs a qualitative methodology grounded in interpretive paradigms, utilizing a single case study of Lai Chi Wo, a traditional Hakka village in Hong Kong. Data are gathered through semi-structured interviews, field notes from observations, and secondary sources. Semi-structured interviews are conducted with purposively selected informants, including local community members, agriculture and tourism practitioners, and policy stakeholders. Theoretically, this study extends the SLA to the context of agricultural tourism, which provides a more comprehensive tool for evaluating the role of agricultural tourism in fostering livelihood sustainability. Practically, it underscores the potential of SLA-informed agricultural tourism to strengthen rural resilience through income diversification, cultural identity preservation, and biodiversity conservation, while also reimagining Hong Kong’s tourism landscape.
Project Reference No.: UGC/FDS16/E28/25
Project Title: Sustainable Development of Photovoltaic (PV) Panel Disposal for Green Concrete Production towards Economic, Energy and Environmental Benefits
Principal Investigator: Dr LAM Stephen Siu-kei (HKMU)
Abstract
The increasing adoption of photovoltaic (PV) panels since the first scaled implementation by the Hong Kong Government in 2005 has led to growing concerns regarding their disposal as they reach the end of their 20-25 year lifespan. The imminent influx of PV panel waste poses significant environmental challenges due to the presence of toxic materials such as lead and cadmium, which can contaminate soil and groundwater if improperly disposed in landfills. Concurrently, the construction industry faces a shortage of river sand, a critical component in concrete production. This research proposal aims to address these dual challenges by recycling PV panel disposal into green concrete, providing environmentally sustainable and economically viable solutions.
The Hong Kong Government has made considerable efforts towards sustainable development to restrain the rise in energy demand while maintaining a cleaner environment. Since 2005, PV panels have been implemented at various scales in Hong Kong. However, with their lifespan being within 25 years, a substantial amount of PV panel waste is expected to emerge by 2030. Currently, disposal options are limited to landfilling, which poses environmental risks due leaching of toxic and heavy elements. Despite the growing concern, there are limited research regarding the recycling applications and its environmental impact if incorporating PV panel disposal into concrete. This project aims to fill in this gap by investigating the feasibility of recycling PV panel disposal for societal use rather than allowing it to end up in landfills and pollute the environment.
PV panels are made of crystalline silicon extracted from sand, in theory, making them a potential substitute for river sand in concrete production. According to the Chief Executive’s 2023 policy address, numerous new building and construction projects are planned in this decade, significantly increasing the demand for construction concrete. As Hong Kong relies on foreign supply for river sand, limited availability could impact costs and burden overall construction budgets. Recycling PV panel disposal as a substitute for river sand could become a viable option.
This project aims to contribute to environmental conservation, resource efficiency, and the advancement of green building practices. This study will comprehensively evaluate different properties of green concrete produced with varying mix ratios and size of PV panel fragmentary. The research will be conducted in several phases, including processing to formulate concrete mixtures, rigorous testing and site application. Additionally, the project will explore the environmental implications and economic viability of using recycled PV materials in concrete. The anticipated outcomes include developing and creating green concrete and establishing guidelines for its safe and effective use in construction.
Project Reference No.: UGC/FDS16/M03/25
Project Title: Investigation of Antitumor Effects of Vitamin D in Giant Cell Tumor of Bone
Principal Investigator: Dr LAU Carol Po-ying (HKMU)
Abstract
Giant Cell Tumor of Bone (GCTB) is classified as an intermediate malignant bone tumor, which is more common in Asia than Western countries and contributes around 20% of the primary bone tumors. GCTB has high postoperative recurrence potential. Adjuvants are frequently used along with surgeries to lower local recurrence of the cancer; nitrogen-containing bisphosphonates and denosumab are the most common adjuvants utilized in GCTB patients. Nitrogen-containing bisphosphonate (zoledronic acid) therapy combined with surgery (curettage or en bloc resection) has been demonstrated to reduce postoperative recurrence rates. However, there are concerns such as necrosis of the jaw bones and abnormal bone fractures with the long‐term use of bisphosphonates. Furthermore, many recent studies have shown that preoperative denosumab therapy raises the relapses in patients. Denosumab induces abnormal hardening of the local bone, and this makes it very difficult to remove tumor cells completely as they may be entrapped within the thickened new bone. Due to the various side effects of the current adjuvants, there is an urgent need to develop a novel adjuvant for treating GCTB.
In our previous study, we demonstrated that simvastatin inhibited cell viability through induction of cell apoptosis and inhibition of cell proliferation. Furthermore, simvastatin stimulated osteogenic differentiation of the neoplastic stromal cells into their terminal status, which paused the proliferation of the tumor cells. The results from our RNA-sequencing showed that simvastatin-mediated osteogenic differentiation of GCTB stromal cells is through the activation of the vitamin D signaling pathway. In fact, many studies have demonstrated that the hormonally active form of vitamin D, 1α,25-(OH)2D3, known as calcitriol, inhibits proliferation and induces differentiation of cancer cells. Moreover, it can stimulate cell apoptosis and hamper angiogenesis in the tumor microenvironment. Therefore, it is worthwhile to investigate how the activation of the vitamin D signaling pathway regulates osteoblastic differentiation of GCTB stromal cells, as shown in our previous study, which may lead to the development of a novel adjuvant for curing the disease.
In this study, we will use both in vitro and animal models to investigate the antitumor effects of vitamin D3 in GCTB. Through the thorough investigation of the vitamin D signaling pathway in GCTB stromal cells, this study will help us to understand the underlying molecular mechanisms of the antitumor effects of vitamin D3. This will shed light on the development of specific GCTB adjuvants, which may reduce recurrence rates, economic burden in patients, and societal health care resource utilization.
Project Reference No.: UGC/FDS15/H26/25
Project Title: After fighting cancer: How do breast cancer (BC) and gynaecological cancer (GC) survivors negotiate and reconstruct selfhood in intimate relationship and the workplace in Hong Kong
Principal Investigator: Dr LAU Flora Pui-yan (Shue Yan)
Abstract
Conventional studies on breast cancer (BC) and gynecological cancer (GC) survivors primarily focus on psychological, sexual, relational, and physical stress, as well as social stigma. However, beyond empirical analyses of these survivors' experiences, there is a need for sociological insights to theorize their post-recovery conditions. In addition, while the social science literature on BC and GC survivors largely centers on Western countries and the Middle East, research on these survivors in Hong Kong has been limited. Among the few studies conducted in Hong Kong, the focus has been predominantly on emotional well-being and quality of life (e.g., Chow et al., 2018; So et al., 2014; Zeng et al., 2011), as well as work productivity (e.g., Cheng et al., 2023; So et al., 2022), indicating significant opportunities for further exploration.
This research aims to explore how breast cancer (BC) and gynecological cancer (GC) impact the sense of femininity in survivors and examine whether and in what ways their experiences have shaped their self-identity. It will discuss sex and intimacy, as well as the working life of BC and GC survivors, as vital domains of their private and public life respectively. Erving Goffman’s negotiation of self (1961) and the feminist orientation of feminine and sexual ideals will serve as the theoretical entry point. The core questions under study include: How does the loss or impairment of female organs affect femininity and, consequently, the sexual and intimate relationships of BC and GC survivors? How does cancer affect their work life and career aspirations? How do they cope with their transformative experiences, and what practical and humanistic strategies can aid in navigating post-recovery life? This study will advocate for the reclamation of selfhood by understanding how negative influences arise, thus purposively reconstructing the self. Additionally, recognizing the (often unspoken) challenges faced by BC and GC survivors, the proposed study will explore appropriate service and policy interventions for local Non-Governmental Organizations (NGOs) and the Hospital Authority (HA) to enhance the well-being of this population.
This study will utilize qualitative in-depth interviews, conducting a total of 80 interviews with a diverse range of subjects. Participants will include 40 BC survivors, 15 GC survivors, 10 intimate partners and coworkers, and 15 professionals from the allied medical and healthcare sector, such as medical social workers, counselors, practitioners from healthcare NGOs, hospital volunteers, and nurses. To ensure a diversity of voices and experiences, there will be no restrictions on the racial and cultural backgrounds of interviewees, provided that they have been living and working in Hong Kong at the time of the interview. However, a minimum of 45 local Chinese survivors is anticipated among the total participants (out of 55 survivors).
This study is expected to make several significant contributions to existing research. First, it will introduce the themes of sex and intimacy, femininity, and selfhood—areas that are rarely studied and yet are crucial to local cancer research. The findings regarding the sexual and intimate experiences of BC and GC survivors will provide alternative insights into self and identity analysis. Second, the results of this study will assist managers and practitioners in the medical and healthcare sectors, including the Hospital Authority (HA) and NGOs, in designing effective strategies and services to support these survivors. Moreover, it will raise the awareness among human resource management in the job market about the conditions of female cancer survivors with a humanistic approach.
Project Reference No.: UGC/FDS24/H17/25
Project Title: Pathways to Sustainable Consumption Behaviour: Using the Power of AR and VR Technologies in Driving Circular Movements in Hong Kong
Principal Investigator: Dr LAU Mei-mei (PolyU SPEED)
Abstract
The Hong Kong Government has promoted public education campaigns like “Dump Less, Save More, Recycle Right” (EPD, 2018) to encourage recycling. This is one of the R concepts of the circular economy - reducing, reusing, and recycling (3Rs) as key strategies to address climate change and environmental pollution. In particular, recycling plays an important role in reducing natural resource consumption, making better use of recyclable materials, and reducing environmental pollution. Even though the Government has been developing educational programmes on recycling, it is unclear about how much such efforts have influenced teenagers to adopt environmentally friendly behaviours and turn their good intentions into long-term sustainable habits, as stated in the common “attitude-behavior gap”. This study examines how immersive technologies, augmented reality (AR) and virtual reality (VR) focusing on recycling and waste reduction, that help motivate teenagers in Hong Kong to adopt environmentally sustainable behaviors. To address these gaps, this study adopts the stimulus-organism-response (S-O-R) framework, integrating innovation diffusion theory and customer value theory to explore how AR and VR can motivate behavioral change and habit formation.
A mixed-methods approach is used to investigate how immersive and gamified experiences influence teenagers’ perceptions of value, attitudes toward sustainability, and their ability to adopt recycling behaviors. The first study will examine how AR/VR games influence teenagers’ attitudes and behaviors toward sustainability and how personal values and attitudes affect their responses. The second study will involve a follow up investigation of participants over three months to see if their sustainable habits improve and persist. Our findings will provide insights into how immersive technologies can help transform teenagers’ attitude into actual sustainable behavior. Through the use of cutting-edge AR and VR tools, the study contributes to advancing circular economy principles, reducing waste, and achieving global sustainability goals. Finally, practical insights will be provided to educators, policymakers, and organizations to design more effective programs and promote recycling and waste reduction, ultimately contributing to a greener Hong Kong.
Project Reference No.: UGC/FDS13/E01/25
Project Title: Assessing Spatiotemporal Variation of Air Pollution in Public Transport Facilities using an Integrated IoT-GIS-based Sensor Monitoring Network in Hong Kong
Principal Investigator: Dr LEE Celine Siu-lan (Chu Hai)
Abstract
Air pollution has raised profound public health concerns as it leads to a large number of premature deaths globally each year and causes substantial harm to human health. A vast number of public commuters are exposed daily to air pollutants inside public transport interchanges (PTIs). PTIs are unique microenvironments that were not adequately addressed in recent air quality studies, as these have mainly focused on indoor or outdoor environments, and thus there is limited information on the air quality in PTIs and transport vehicles. Meanwhile, the spatial variability of air pollutants inside PTIs, which is crucial in predicting their pollution levels remains unclear.
In this research study, we aim to conduct intensive air sampling in representative semi-enclosed PTIs in highly urbanized areas of Hong Kong in two seasons (i.e., summertime and wintertime). The profile of the spatio-temporal trends of multiple air pollutants will be analyzed using a combination of Internet of Things (IoT) and geographical information system (GIS) techniques. We will assess the health risks associated with these air pollutants using the health indices. A receptor model was applied to the data for the apportionment of pollution sources of air pollutants in the microenvironment. These findings will further help in highlighting the key factors that influence the air quality inside PTIs and practical measures in controlling the associated health risks. The results of this proposed study will illuminate the air pollutants present in ambient air and their interactions with humans. Our objectives are to (1) develop and apply IoT-based sensors for a monitoring network of multiple air pollutants at representative PTIs; (2) elucidate the spatial variability of multiple air pollutants and the best fitted interpolation models; and (3) develop a Web-based alarm system to inform the public on the associated health risks and to better manage the air quality inside the PTIs. An improved understanding of these issues will contribute to the efforts to mitigate the sources and risks of toxic air pollutants and address public health challenges from air pollution, and provide scientific evidence for relevant departments to formulate more targeted air quality improvement measures.
Project Reference No.: UGC/FDS41/H08/25
Project Title: The impact of parental stress, parenting practices, and the provision of learning environments on the development of working memory in kindergarten-aged children
Principal Investigator: Prof LEE Kerry (YCCECE)
Abstract
Working memory capacity (WM) affects the efficiency with which higher cognitive processes can be carried out and is strongly associated with academic performance. Although WM increases significantly from early childhood to adolescence, Lee and Bull (2016) found that WM grew at similar rates across individuals. Furthermore, they found significant WM differences even amongst kindergarten-aged children. These findings mean that children with low WM in early childhood tend not to benefit from faster WM growth. Instead, their disadvantage persists into adolescence. If we are to design early WM interventions to assist these children, ascertaining the factors and mechanisms that produce individual differences in WM is vital.
Previous studies suggest that socioeconomic status (SES), parental stress, parenting practices, and parents' provision of learning opportunities are associated with children's WM development (Duran et al., 2020; Lawson et al., 2018). However, questions remain on the interrelations between these factors and the mechanism through which they influence WM. Regarding the interrelations between these variables, we will examine the relative importance of two mechanisms. First, the family stress model suggests that WM is influenced by parenting practices, which are themselves influenced by parental stress and SES. Second, the family investment model suggests that WM is mainly influenced by whether families have the financial resources to provide appropriate learning opportunities. We will also examine which parenting practices are more closely related to WM development. Based on findings from a pilot study (Lee, 2023), we expect children with lower WM to have parents who are inconsistent in their parenting practices or who are permissive and provide little guidance.
The study will use a longitudinal cross-panel design to track the development of 4 to 5-year-olds (N=191) over 14 months. Children will be administered, across four occasions, three WM tasks (Lee, 2023): Corsi, backward digit span, and Animal Updating. The children will be recruited from low-, middle-income, and high-income kindergartens. Their parents will be asked about their socioeconomic status, worries about financial sufficiency, stress, and parenting style. They will also be asked about their investment in their children's learning. We will use a statistical model that disaggregates within-person versus between-person variation (Hamaker et al., 2015) to test the interrelations between the various constructs. This study will provide information on the specific parents-related variables and the pathways through which they influence WM development: information vital for the development of new WM interventions.
Project Reference No.: UGC/FDS16/H05/25
Project Title: In the Perspective of Recorder: A Research on the Authoring and Recording of Ancient Chinese Etiquette Texts
Principal Investigator: Dr LEE Lok-man (HKMU)
Abstract
Rituals are the core of Chinese culture, and the important ancient rituals of China have been recorded in ancient Chinese ritual texts and being preserved to the present day. As one of the Confucian classics, the Yili (儀禮Etiquette and Ceremonial) is also the origin of ancient Chinese ritual texts. Unlike other classical texts, ”rituals” often involve the practice and demonstration by the performers. Scholars must imagine the actual execution of these rituals based on textual records to reconstruct the ritual procedures of the time. In other words, understanding the methods of recording and compiling texts becomes the key to interpreting the Yili text.
This research project, spanning two years, aims to conduct a systematic and in-depth study on the Yili text from the perspective of the recorder. Although there has been no shortage of research on the Yili in academia, few studies have explored its recording and compilation methods from the recorder's perspective and responded to academic discussions. Therefore, the principal investigator hopes to use innovative methods, combined with past research experience, to address three research questions in this project: (1) What techniques did the recorders use to document the actual ritual details into ritual texts? (2) Are the writing styles of the seventeen chapters of the Yili consistent? (3) How can the conclusions of this investigation be reconsidered within the historical context of Chinese ritual studies, especially regarding the textual integrity and compilation of the Yili, and how do they respond to related academic discussions?
Regarding the design and methodology of this study, the PI aims to conduct detailed textual analysis in a systematic and organized way. The research will be divided into two main stages: (1) Data organization and classification. The research team will classify the transmitted Yili texts, supplemented by the excavated version, into different categories based on their nature. Additionally, the team will make use of key glossaries of the Yili text, in order to thoroughly integrate the Yili texts for reference and comparison. (2) Textual analysis. Based on the data organized in the first stage, the research team will analyze each entry using an intertextual research method to identify and summarize the various recording patterns and techniques of the Yili recorders, while also attempting to identify any special or unusual writing styles. Furthermore, the team will analyze and organize the chapters to depict the focus or characteristics of each chapter's recording. The findings will be compiled into papers and published in recognized academic journals.
This study, through a new perspective on the writing of the Yili, aims to summarize the recording methods of ancient ritual texts, filling academic gaps and opening up new research methods for Confucian classics. Moreover, this research project will help to re-understand the Yili, reduce reading difficulties, and thereby enhance the importance of ritual text education in universities. In the long run, the project will also contribute to the inheritance and dissemination of Chinese rituals, providing conditions and a solid academic foundation for the restoration and construction of Chinese rituals, and aiding in the preservation and dissemination of Chinese civilization.
Project Reference No.: UGC/FDS15/B01/25
Project Title: Examining Impacts of Channel Addition/Elimination on Perceived Control and Concern about Cybersecurity: A Scenario-based Survey Study of Virtual Banking
Principal Investigator: Dr LEE Philip Tin-yun (Shue Yan)
Abstract
Due to the Covid-19 pandemic, the elimination of brick-and-mortar channels has gained much attention in recent years. Previous studies on retailing and services have mainly focused on channel addition and integration. However, relatively few studies have examined channel elimination and forced adoption of remaining channels. Specifically, the relationship between the distribution of offline touchpoints and the cybersecurity of online channels has not been examined.
In this study, we will conduct scenario-based surveys to understand the influence of channel elimination and addition, especially in the marginal cases of adding the first touchpoint and removing the last touchpoint of offline channels, on people’s perceived control and concern about the cybersecurity of online channels. Highlighting the difference between touchpoints and channels, we define that a channel choice is eliminated when all touchpoints of a channel type are removed. The survey scenarios will be contextualized in the banking sector in Hong Kong. In the first study, we will demonstrate that eliminating all, but not some, brick-and-mortar touchpoints from multichannel, traditional bank will result in a significant reduction of perceived control and a significant increase of concern about cybersecurity. In the second study, we aim to find that by adding merely one single brick-and-mortar touchpoint to a digital-only virtual bank can significantly increase people’s perceived control of the bank’s online channels. The touchpoint addition will also significantly reduce people’s concern about the cybersecurity of its online channels.
Regarding theoretical implications, this study will establish relationships between offline touchpoints and online cybersecurity. It will also address the current research gap of overlooking the gradual process toward complete channel elimination. In terms of practical implications, we recommend that multichannel businesses avoid closing all offline channel touchpoints. Negative effects of perceived control due to the deprived freedom of channel choice are substantial when the size of the brick-and-mortal channel is reduced to none. If a multichannel business decides not to keep even one brick-and-mortar store, we will recommend the business be prepared to upgrade its cybersecurity of online channels.
Furthermore, for digital-only businesses, we will propose an additional approach to alleviate cybersecurity challenges alongside efforts on online channels: adding a brick-and-mortar touchpoint which leaves customers an extra channel choice to complete their customer journey. Customers may be less hesitant to make their online purchases due to the existence of an offline touchpoint.
Project Reference No.: UGC/FDS16/H34/25
Project Title: Edmund Blunden in Hong Kong, 1953-1964
Principal Investigator: Dr LEE Sarah Sze-wah (HKMU)
Abstract
Edmund Blunden (1896-1974) was a British poet who was well-known for his war poetry written as a soldier and veteran of World War I. He subsequently had a long and prolific global career of writing and teaching English literature, including as Professor of English at the University of Hong Kong (HKU) (1953-1964), a period which entailed critical development of English studies at the university and the post-war modernization process of the city. Blunden’s poetic output in Hong Kong and his decade-long career and life here have however been little studied, accounting for only 15 pages in the only biography to date. This project seeks to fill this gap by producing a comprehensive account of Blunden’s life in Hong Kong which charts his legacies and contributions to the city as well as global Anglophone modernism, primarily through archival and oral history methods.
Blunden’s involvement in East-West literary and cultural exchange in Hong Kong during the Cold War period is a significant literary and historical case study. Firstly, he was one of the pioneering Anglophone writers to take Hong Kong as subject matter, setting the precedent of Hong Kong writing in English. Secondly, as a poet, professor and scholar of English literature, his expertise was not only manifested in the curriculum and teaching at the English Department, but also in drama education at HKU, contributing to the nurturing of a new generation of talents instrumental to further developments in Anglophone writing and drama in Hong Kong. Thirdly, his public presence as guest speaker at various educational institutions and arts societies was ubiquitous and conspicuous, creating wider influence beyond the HKU in the arts and culture. Lastly, his international network extends his influence to regions outside of Hong Kong, and this project will examine his connections with two nearby locations: Japan in which he previously worked twice for a total of six years and remained in active correspondence with many individuals, and more significantly China, to where he made two visits in 1955 and 1964 respectively, and was received by the Prime Minister Chou En-lai on the former visit with a delegation from the HKU.
This project aims to investigate Blunden’s contributions to and legacies in Hong Kong, in terms of Anglophone writing and literary criticism, the teaching and learning of English literature, and the promotion of arts and culture, not only in the local context but also with wider impacts in light of his global network. The investigation will be conducted with an innovative multi-pronged approach, especially by using firsthand data from undiscussed archival materials (including a large volume of correspondence), and the conducting of oral history interviews with his former students at HKU, his family, and others who knew him during the period. In addition, his creative and scholarly outputs from Hong Kong will also be analysed in relation to critical literature on global modernism and the Cold War context.
Not only will this qualitative study fill existing gaps in Blunden’s biography, but it will also enrich the understanding of the development of Anglophone literature, drama, and the higher education in Hong Kong in the 1950s-1960s and the legacies to the present day. Moreover, it will put forward Hong Kong’s contribution to East-West literary and cultural exchange, transnational modernism and global Anglophone literature in the 20th century. This interdisciplinary project will result in a range of research outputs, including a scholarly monograph of Blunden in Hong Kong, journal articles, and conference papers. Public impact will be generated through the development of an interactive website that details Blunden’s life in Hong Kong comprehensively and a chatbot on his poetry written in Hong Kong, both of which aim to provide scholarly-generated knowledge to the public in an accessible and engaging manner to further the understanding of Blunden and his contexts.
Project Reference No.: UGC/FDS16/H27/25
Project Title: What L2 learners ‘know’ and how they ‘feel’ about L2 idioms: A cross-linguistic, psycholinguistic and affective study of L2 idiom processing for Chinese L2 learners of English
Principal Investigator: Dr LEUNG Chung-hong (HKMU)
Abstract
Idioms such as fall head over heels and see red are fixed expressions whose figurative meanings are distinct from the meanings of their constituent words. While idioms constitute a significant portion of everyday language use among native (L1) English speakers, second language (L2) learners of English face challenges in comprehending and producing idiomatic expressions. Research highlights the lower frequency and accuracy of L2 learners in using English idioms compared to L1 speakers. Cross-linguistic studies attribute these difficulties to L1-L2 transfer effects, and psycholinguistic research also suggests systematic differences in idiom comprehension between L1 speakers and L2 learners.
In recent years, there has been a growing interest in exploring how psycholinguistic properties of idioms (i.e., what people ‘know’ about idioms) and their affective properties (i.e., how people ‘feel’ about idioms) influence idiom processing. While these properties have been extensively studied in L1 idiom processing, their impact on L2 idiom processing remains underexplored. In other words, investigating what L2 learners 'know' about English idioms (e.g., how they associate fall head over heels with the meaning ‘deeply in love with’) and how L2 learners ‘feel’ about English idioms (e.g., whether see red meaning ‘very angry’ makes L2 learners feel intense) will enrich the body of knowledge regarding L2 idiom processing and will better inform L2 English teachers and students of their teaching and learning practices concerning English idioms.
This project aims to adopt a cross-linguistic framework to investigate L2 idiom processing in Chinese L2 learners of English by collecting and analysing their normative ratings and responses to the psycholinguistic and affective properties of English idioms. The project will be structured across seven phases spanning a 24-month period. In Phase I, 250 English idioms will be meticulously selected for the study based on specific criteria which include widespread usage and representation from five source domains. Subsequently, a unique cross-linguistic comparative model will be deployed in Phase II and the selected English idioms will be categorized into various types of correspondences with reference to their linguistic and conceptual overlaps with the Chinese equivalents. The project will then progress to Phases III to V which involve the development of an online questionnaire on the Qualtrics platform and the dissemination of the questionnaire to 400 Chinese L2 learners of English to collect their ratings and responses to the psycholinguistic and affective properties of the selected English idioms. In Phase VI, the data collected from the online questionnaires will be analysed through rigorous processes including data trimming, coding, and correlation measures. The culmination of the project will take place in Phase VII which involves incorporating the findings of the project into the Affective and Psycholinguistic Ratings of English Idioms List (APREIL), alongside paper presentations and journal articles to share the comprehensive insights gained from the research project.
Theoretical implications from the project are significant as it aims to delve into L2 learners’ knowledge and feeling of English idioms within the cross-linguistic context, which will shed light on the intricacies of L2 idiom processing and will inform language education practices. Pedagogically, the outcomes of this project will hold promise for enhancing idiom teaching and learning strategies, figurative language curriculum development, and figurative competence among L2 learners.
Project Reference No.: UGC/FDS24/B18/25
Project Title: From Play to Health: Enhancing Elderly Engagement with Exergames Through Affordance and Goal Framing Theories – A Longitudinal Study and Experiments in Hong Kong
Principal Investigator: Dr LEUNG Wilson Ka-shing (PolyU SPEED)
Abstract
Hong Kong is experiencing one of the fastest rates of population aging globally, with the elderly population expected to peak in the coming decade. The number of individuals aged 65 and older is projected to increase from 1.5 million in 2021 to 2.52 million by 2039 (Health Bureau, 2023). One of the most significant concerns associated with aging is the increased risk of falls, which is often attributed to muscle loss and weakness, compounded by a lack of regular exercise, which can lead to fractures that require prolonged and challenging recovery times; in some cases, they can lead to coma and even death (Chaabene et al., 2021). This situation places a considerable burden on the local public healthcare system. One possible solution to reduce health decline in older adults is to engage in regular exercise. In view of this, the Hong Kong government keeps allocating resources to encourage the elderly to exercise more, sponsoring district and neighborhood community centers to provide exercise training and equipment for seniors (Social Welfare Department, 2025). Despite the well-known positive effects of exercise, the physical activity levels of older adults in Hong Kong are even worse than they were before the COVID-19 pandemic (Wang et al., 2023). Consequently, the health condition of elderly individuals will continue to worsen if they do not actively engage in regular exercise.
One type of gerontechnology, known as exergames, has recently emerged for seniors. Exergame, a term that combines “exercise” and “gaming”, is defined as a form of physical activity that integrates various game elements into exercise routines. Exergames motivate users to engage in physical activity, including activities that enhance strength, balance, and flexibility, through an appealing and interactive format (Vernadakis et al., 2015). Players can receive immediate feedback about their exercise performance from the game system (Bakker et al., 2020). This interesting exercise method has great potential to promote healthier aging and help reduce government expenditures in the long run. However, empirical studies on how exergames can effectively encourage older adults to exercise regularly and enhance their well-being remain limited. Drawing on our expertise in healthcare technology, our interdisciplinary research team, composed of information systems (IS) and healthcare technology researchers, is collaborating with both elderly service organisations (e.g., Sik Sik Yuen) and a gerontechnology company specialising in exergames (e.g., Medmind Technology Limited). Through three studies, we aim to investigate how elderly users engage with exergames, focusing on the interplay between user goals and game features as framed by affordance theory and gamification elements, and how this engagement subsequently impacts their usage behaviors and well-being.
To address the identified research gaps, we developed a three-stage theoretical model that integrates affordance theory, principles of physical training techniques, and self-regulation theory. This model aims to explain how elderly users engage with exergames by leveraging the game's affordances in pursuit of their personal goals (e.g., well-being). With the assistance of the above collaborators, our multistage model will be examined by recruiting elderly people from their elderly centres. Mixed-methods design will be applied to develop and validate new measurements. Additionally, two experiments will be conducted to investigate the effectiveness of gamification elements in encouraging elderly users to participate in exercise. PLM-SEM and ANOVA will be used for our data analysis. We will conduct a qualitative study to augment the insights gained from the survey results. Overall, our findings not only contribute to IS literature, but also provide managerial guidance for the government, elderly service organisations, the gerontech companies, and tertiary education about how to enhance the physical health among the elderly effectively.
Project Reference No.: UGC/FDS11/E08/25
Project Title: Developing an Efficient Animation Production Pipeline via Deep Production Material Understanding and Motion Analysis
Principal Investigator: Dr LI Chengze (SFU)
Abstract
Traditional animation production demands great efforts in hand-drawn steps per scene, so only big studios can afford it. This project builds practical artificial intelligence (AI) tools that speed up the same workflow artists already use, while keeping creative control in their hands. First, we will teach AI to parse rough storyboards (which are the sketch pages that outline each shot), so it can recognize characters, camera moves and directors’ notes. Next, the system will turn that understanding into “animatics”, quick motion previews used to plan timing and scene changes. We then use AI to help color and shade key frames and to synthesize smooth motion between them, with easy controls for directors to tweak poses, objects and cameras. Alongside the tools, we will assemble and release a large, well-labeled dataset that links storyboards, animatics and finished clips, so future teams can build on our work. The result is an interactive, artist-in-the-loop platform that lowers cost and shortens schedules without sacrificing style or quality. This will help independent creators and small studios produce more ambitious work, and it can also improve film pre-production, where storyboards and animatics are standard. We will share results openly through papers, code and models to accelerate adoption across the creative community around the world.
Project Reference No.: UGC/FDS14/B09/25
Project Title: Monetary Incentives and Creativity: An Experimental Investigation
Principal Investigator: Dr LI King-king (HSUHK)
Abstract
Creativity is an engine for firm’s innovation, business strategy, economics development, and for making a better society (Charness and Grieco, 2023; Aghion and Howitt, 2008; Ko and Butler, 2007). Understanding how to foster creativity is an important question not only for academic research but also great social value. Currently, there is limited investigation into this issue, and the evidence is both sparse and mixed.
This study experimentally investigates the effect of monetary incentive on creativity using laboratory and field experiment. We experimentally investigate whether there is crowding out effect of monetary payment on creativity. In particular, we compare the creativity of subjects when there is monetary incentive for creativity compared to when there is no monetary incentive. We experimentally compare the performance of payment schemes under different types of creativity task including writing and designing a new product. We conduct field experiment to investigate the effect of different payment schemes. This will be one of the first field experiment on creativity.
Our findings will have important policy implications for the business strategies of companies in designing optimal incentive schemes to foster creativity and innovation, which are vital for companies’ success in a world where innovation is key to driving business and maintaining competitiveness in the economy. Our findings will also have important implications for education policy in designing optimal ways to nurture students' creativity. For example, designing optimal reward schemes in STEM competitions.
Project Reference No.: UGC/FDS14/E04/25
Project Title: Towards Optimizing Generative AI Inference in Resource-Constrained Edge Environments
Principal Investigator: Dr LI Tan (HSUHK)
Abstract
Large Language Models (LLMs) and multimodal generative artificial intelligence (GenAI) are transforming how we approach various tasks, from document summarization and visual scene understanding to real-time language translation and interactive dialogue systems. Traditional GenAI services rely on cloud-based inference, where user tasks (prompts) are uploaded to centralized cloud servers for processing. However, as these GenAI applications expand into mobile and distributed scenarios such as smart homes and autonomous vehicles, cloud-based inference faces growing challenges in meeting latency and cost requirements. Edge inference, which deploys LLMs on distributed edge clusters, has emerged as a promising solution to ensure responsive and efficient service delivery. This shift towards edge inference is crucial for making GenAI services more accessible and practical for end-users.
However, the current edge inference framework for GenAI tasks faces three critical challenges due to computation and communication resource constraints. First, GenAI prompts submitted by users vary widely in input sizes, modalities (e.g., text, images), and latency requirements. Meanwhile, edge clusters differ significantly in their GPU memory, computational capabilities, deployed models, and network bandwidth. This real-world dual heterogeneity between user demands and edge resources is often overlooked by existing prompt scheduling schemes. As a result, prompts are frequently assigned to unsuitable edge clusters, causing system load imbalance. This leads to degraded performance in terms of long response times and poor inference quality. Second, processing multi-modal GenAI prompts, including high-resolution images, lengthy text sequences, or video data, under bandwidth constraints introduces substantial end-to-end latency. This latency arises from both the prompt uploading and the inference phase. Such bottlenecks are especially critical for latency-sensitive applications, such as real-time traffic video analysis or interactive AI assistants, where delays can severely impact user experience. Finally, there is a lack of specialized platforms for evaluating the performance of edge inference frameworks for GenAI tasks. This makes it difficult to systematically validate and benchmark optimization algorithms.
In response to these three challenges, we propose an optimized GenAI edge inference framework under resource-constrained environment. Our framework addresses the identified challenges through three key components: 1) An intelligent task- and load-aware scheduling scheme that effectively manages the dual heterogeneity between user prompts and edge clusters, achieving balanced system load while optimizing user quality of service; 2) A latency- and modality-aware prompt compression mechanism for multi-modal GenAI tasks that minimizes inference quality degradation while meeting latency constraints under bandwidth-limited conditions, and 3) A comprehensive evaluation platform that enables systematic benchmarking of edge GenAI inference frameworks under diverse workloads and network conditions.
This project, if successful, will significantly enhance the efficiency of edge inference for real-world GenAI tasks by optimizing user satisfaction and resource utilization. Moreover, it has the potential to seamlessly integrate with various existing edge intelligence architectures, such as smart manufacturing and intelligent transportation systems, thereby improving the operational efficiency of urban environments and enhancing the quality of life for citizens.
Project Reference No.: UGC/FDS16/E24/25
Project Title: Sustainable Maintenance of Highway Bridges Considering Climate-Driven Deterioration under Traffic Demands
Principal Investigator: Dr LI Yaohan (HKMU)
Abstract
Highway bridges are essential components of transportation infrastructure, particularly in densely populated regions like the Greater Bay Area. However, these structures face growing challenges due to the combined effects of climate-driven deterioration and increasing traffic demands. Over time, aging materials and environmental factors, such as temperature fluctuations, humidity variations, and chloride-induced corrosion in coastal regions, accelerate the deterioration of structural resistance. Meanwhile, the continuous rise in traffic demands may further exacerbate structural deterioration. The combined impact of these influences poses a significant risk to structural safety and long-term performance, necessitating life-cycle maintenance for coastal highway bridges.
Despite these challenges, current approaches mainly focus exclusively on either climate-induced deterioration modeling or short-term traffic loading impacts, without adequately addressing their combined, time-dependent effects on structural resistance loss. In particular, existing traffic models have not captured spatial and temporal variability in extreme loading scenarios, while the deterioration models of reinforced concrete structures lack sufficient experimental validation under local climate conditions. Additionally, with the increasing global emphasis on carbon neutrality, there is a pressing need for sustainable maintenance strategies that balance structural performance, economic feasibility, and environmental impact over the bridge lifetime.
This project aims to bridge these gaps by developing a multi-disciplinary framework for evaluating and managing the life-cycle maintenance of RC highway bridges, under the combined influence of long-term climate-driven deterioration and traffic demands. The proposed framework will integrate stochastic traffic load modeling, climate-specific resistance degradation assessment, AI-aided reliability analysis, and cost-effective and sustainable maintenance strategies. By incorporating carbon footprint assessments into maintenance planning, this project aligns with global sustainability goals and supports the local carbon neutrality targets of Hong Kong and the Greater Bay Area. The findings will contribute to the resource-efficient infrastructure management strategies, ensuring long-term safety, functionality, and sustainability.
Project Reference No.: UGC/FDS16/E25/25
Project Title: From Language to Logic: Intelligent In-Context Learning for Software Modeling Automation
Principal Investigator: Dr LI Yishu (HKMU)
Abstract
Software modeling, a critical step in software development, involves translating Natural Language (NL) requirements into Unified Modeling Language (UML) diagrams to ensure high-quality code implementation. Despite advances in Natural Language Processing (NLP), Large Language Models (LLMs) face significant obstacles in handling complex software modeling tasks, such as generating class-level designs that require a deep contextual understanding of NL requirements. Recent research has investigated the potential of LLMs in performing initial software modeling tasks, such as deriving conceptual class diagrams. Current approaches, including single-pass prompting and Chain-of-Thought techniques, fall short of capturing the contextual relationships necessary for comprehensive software modeling since it is a complex task necessitating comprehensive analytical and collaborative efforts in analyzing NL requirements.
In-context learning (ICL) has demonstrated the ability of LLMs to address benchmark tasks using training examples (i.e., few-shot learning). It has produced promising results in NLP tasks that often require complex reasoning. To address the identified limitations, we propose ICLMod, an intelligent ICL framework designed to simulate the human analytical and collaborative process in software modeling. Unlike traditional prompting methods, ICLMod incorporates advanced ICL techniques, such as role-playing and tree-of-thought, to guide LLMs in generating contextually relevant and logically consistent UML diagrams. This framework sequentially addresses key software modeling tasks, including class diagrams, sequence diagrams, activity diagrams, and state diagrams. At each stage, ICLMod provides contextual input to refine LLM outputs, ultimately translating from language (i.e., NL requirements) into logic (i.e., UML diagrams), which serve as the foundation for code implementation.
This project will first conduct comprehensive empirical studies on human collaborative patterns and analytical behaviors in software modeling, with a specific focus on industries in the Greater Bay Area (GBA). The insights gained from these empirical findings will be integrated into the advanced prompting techniques, such as role-playing and tree-of-thought, specifically designed for ICLMod. The performance of the framework will be evaluated through a multifaceted approach that combines quantitative metrics (e.g., the completeness and relevance of LLM-generated diagrams compared to human-generated diagrams) and qualitative methods (e.g., detailed case studies and feedback from industry practitioners). This dual evaluation strategy ensures that the outputs of ICLMod are aligned with real-world expectations and industry standards.
To enhance practical applicability, we will develop a tool based on ICLMod that automates the translation of NL requirements into UML diagrams. This tool will be designed to integrate with existing tools capable of converting UML diagrams into code. Consequently, this tool aims to streamline software development, improve efficiency, and bolster the competitiveness of IT industries in Hong Kong and GBA. By addressing the critical gap in NL-to-UML translation, ICLMod represents a transformative step toward intelligent, automated software modeling.
Project Reference No.: UGC/FDS24/E17/25
Project Title: Development of Ultra-Broadband Airflow Transparent Noise Barrier by Combining Meta-Resonance System and Highly Absorptive Units
Principal Investigator: Dr LIANG Shanjun (PolyU SPEED)
Abstract
Residents in Hong Kong are significantly affected by noise pollution in urban areas. According to the Environmental Protection Department (EPD), the primary sources of outdoor noise include construction activities, road traffic, and industrial operations. Noise pollution poses both physical and psychological health risks. Prolonged exposure to noisy environments can lead to hearing loss. Excessive noise levels can also cause stress and anxiety, disrupt communication and concentration, and increase the likelihood of workplace accidents and injuries. Furthermore, noise pollution diminishes the quality of life for residents, negatively impacting the city’s overall liveability. It can also reduce Hong Kong’s attractiveness as a destination for investment and immigration, posing challenges to its long-term development.
To overcome the noise problem, the Government has implemented ordinances and regulations to control the noise source. Meanwhile, noise barriers are extensively used to reduce the noise effects from propagation. However, there is still a large room for improvement for the existing barriers: the conventional mass law requires the barrier to have enough mass density, increasing the weight of the barriers; the airtightness reduces the natural ventilation; for outdoor cases, the wind load increases the expenses on their construction. Also, the residential areas and office buildings sacrifice natural ventilation due to the noise problem, making air conditioning the most significant electricity energy cost in Hong Kong.
The development of manufacturing methodology and wave control science brings new solutions to the noise insulation scheme, primarily to address the conflict between ventilation and soundproofing. The research in recent years shows that the meticulously designed sparse structures reflect the sound yet airflow transparent, which is theoretically possible and experimentally validated. These structures are called acoustic metamaterials or meta-structures, consisting of periodic distribution of functional unit cells. The resonance modes of these unit cells lead to high transmission loss. A broad category of metamaterials’ acoustic behaviour relies on geometrical properties. This property significantly enriches the design pool. From the practical use perspective, these solutions require further improvement in reducing the thickness, improving the performance for low-frequency response and varying noise source positions. Therefore, we propose to enhance the practical application of the airflow transparent noise barrier with theory exploration, numerical study and experimental evaluation in this project. Firstly, based on the acoustic scattering theory, we develop a systematic way to construct effective scattering lattices for broadband noise mitigation with strong reflection and remarkable absorption responsible for low-frequency and high-frequency bands. Secondly, the acoustic characteristics and performance will be measured under impedance tubes and lab-made waveguides. Finally, the developed noise barrier will be experimentally evaluated in practical size under actual application conditions. We will compare our results with the existing solutions in academic and industrial counterparts.
Upon the completion of the project, several designs will be obtained for enhanced ventilation performance and ultra-broadband noise reduction. This will contribute to the development of ventilation noise barriers in academics and applications and promote the widespread use of natural ventilation and soundproofing devices in Hong Kong.
Project Reference No.: UGC/FDS11/E03/25
Project Title: Promoting Zero-Shot Learning Ability for Vision Language Model
Principal Investigator: Dr LIU Hui (SFU)
Abstract
Vision language models (VLMs) like CLIP have shown incredible zero-shot learning ability. For a new learning task, the pre-trained VLMs can perform well without being fine-tuned. Specifically, in the new task, VLMs usually first generate pseudo-labels for samples and then regard those pseudo-labels as supervision signal to guide the prompt learning to make the input suitable for the VLMs, such that the VLMs achieve zero-shot learning on this new task. Although this pipeline has demonstrated effectiveness, it suffers from the following three major limitations. First, the pseudo-labels generated by the VLMs inevitably have labeling errors as the training domain and the downstream domain in the new task is different. The error pseudo-label will degrade the zero-shot learning ability of VLMs. Second, VLMs consist of multiple modalities, and the multimodality gap in the VLMs hinders the reliability of the constructed zero-shot learning model. Finally, the current pipeline is sensitive to the adopted prompt.
In this project, we aim to promote the zero-shot learning ability of VLMs. (1) We will improve the pseudo-labeling strategy. Specifically, instead of selecting a single pseudo label for each instance, we propose to choose a set of candidate pseudo labels for each instance. This strategy makes the correct label more likely to be assigned to the instance but will also bring additional label ambiguity. Therefore, we will formulate this strategy as a multi-objective optimization problem that reduces the candidate label set size and improves the quality of the candidate label simultaneously. Although this strategy can enhance the pseudo-label quality, the ground-truth label may still be located outside the candidate label set (OCLS); we will further investigate how to solve this OCLS problem. (2) We will try to reduce the modality gap in VLMs. First, we will formulate the representation of the embeddings in different modalities as a distribution rather than a point to enhance the robustness of the representation. Then, we will construct a modality-wise classifier to communicate with each other to achieve mutual learning. As different modalities may be skilled in predicting different instances, we will propose an uncertainty-aware classifier fusion strategy to achieve better prediction. (3) We will investigate a more effective and robust zero-shot learning paradigm for VLMs. First, to reduce the influence of pseudo labels, we will leverage self-paced learning to first exploit the easy samples and then include the harder ones to make the zero-shot model more robust. Then, we will ameliorate prompt learning scheme to make the zero-shot model more robust and stable.
This project will benefit the fundamental research on visual language models and machine learning. It will highly reduce the annotation cost for many industrial applications like healthcare and automotive. It will also help local community life related to visual language models like healthcare and education.
Project Reference No.: UGC/FDS25/E04/25
Project Title: Exploration of water/oil separation mechanism from biodegradable poly(3-hydroxybutyrate-co-3-hydroxyvalerate) (PHBV)/poly(caprolactone) (PCL) composite membranes fabricated via electrospinning technology
Principal Investigator: Dr LIU Yaohui (THEi)
Abstract
This research aims to create a new type of environmentally friendly membrane that can separate oil from water, which is especially useful for cleaning up oil spills and treating oily wastewater. The membrane is made from special biodegradable plastics, so it won’t harm the environment after use. The team uses a technique called electrospinning, which produces the mat of extremely fine fibers with tiny porous structure. This structure helps the membrane to absorb oily chemicals while blocking water.
The main materials used are poly (3-hydroxybutyrate-co-3-hydroxyvalerate) (PHBV) and poly (caprolactone) (PCL), both of which can break down naturally over time. By combining these two materials with various ratios, the researchers can adjust how well the membrane repels water and absorbs oil. Early tests show that these membranes can absorb more than ten times their own weight in oil and are very good at keeping water out. By improving the absorption strength and performance, making it a practical and eco-friendly solution to protect the water resources from oil pollution. Through investigating the fundamental science behind the oil absorbing properties, a better product can be achieved based on proven principles. This research could lead to safer, greener ways to protect our water resources from oil pollution.
Project Reference No.: UGC/FDS24/E26/25
Project Title: Reinforcement Learning-Driven Adaptive Traffic Noise Barrier
Principal Investigator: Dr LOH Anthony Wai-keung (PolyU SPEED)
Abstract
Traffic noise has become a significant concern in urban environments, negatively impacting public health and well-being. Conventional noise barriers, while effective to some extent, often require large, bulky structures that are not suitable for low-frequency noise mitigation, especially in space constrained areas. Recent advancements in acoustic metamaterials (AMM) offer promising alternatives due to their unique properties, such as negative mass density and negative bulk modulus, which enable superior sound attenuation at subwavelength scales. This study aims to explore the potential of AMM, focusing on a modified perforated honeycomb-corrugation hybrid (PHCH) metamaterial for low-frequency traffic noise control.
To enhance the adaptability of noise barriers, this research integrates reinforcement learning (RL) and Internet of Things (IoT) technology to develop an intelligent, real-time noise mitigation system. By employing strategically placed sensors along the noise barrier, real-time environmental and acoustic data can be collected and processed through an IoT-based framework, allowing the system to dynamically adjust its acoustic properties in response to changing environmental and noise conditions. RL plays a critical role in this system, optimizing the relationship between the shape parameters of PHCH and their sound absorption performance. By iteratively adapting and improving barrier configurations based on continuous feedback, RL enhances the efficiency of the design process, ensuring that PHCH absorbers meet specific acoustic targets with greater precision and speed.
The research further incorporates magnetorheological elastomers to enable tunable acoustic properties, making noise barriers more effective in dynamic urban environments. The proposed IoT-enabled adaptive noise barrier system leverages active noise control (ANC) techniques, where anti-noise waves are generated to counteract unwanted sound through destructive interference. IoT connectivity ensures that sensor data is processed in real time, allowing for automated barrier adjustments based on predictive noise pattern analysis. This approach significantly enhances the functionality and responsiveness of traffic noise mitigation solutions. Furthermore, the RL system will collaborate with the IoT framework to dynamically adjust the acoustic absorption properties of noise barriers, utilizing real-time environmental and acoustic data.
The AMM prototypes will undergo laboratory testing in anechoic chambers under real-world field environments, with performance evaluated against finite element simulations (FEM). By integrating RL, IoT, and ANC technologies, this research aims to develop a next-generation, intelligent noise control solution that is both space-efficient and highly adaptable. The outcomes of this study have significant implications for urban noise management, particularly in densely populated cities such as Hong Kong, where traditional static noise barriers are often insufficient. This work contributes to the advancement of smart infrastructure and sustainable urban planning, providing a scalable solution to mitigate the growing challenge of traffic noise pollution.
Project Reference No.: UGC/FDS11/H10/25
Project Title: The Importance of Disciplinary Literacy in iGCSE: A Data Driven Study in Hong Kong
Principal Investigator: Dr LOK Beatrice Chak-ying (SFU)
Abstract
Writing skills are crucial for the academic success of all students in school. This research aims to examine disciplinary literacy and writing development during the International General Certificate of Secondary Education (iGCSE) exam years in Hong Kong, and to provide original insights that can benefit teaching and learning in the region. The goal is to explore the language patterns that characterize various disciplinary writings of students across different types of schools in Hong Kong. The present study focuses on analysing student-written texts from subjects taught at iGCSE in participating schools to explore the language features across various disciplines. Of particular, this research focuses on developing discipline-specific high-frequency word lists, grammatical structures, and usage for individual iGCSE subjects. It also aims to identify effective organisational patterns of texts for various types of questions. It is hoped that the data collected from the project can create an authentic data set for teaching, learning, and research purposes. The findings from the study will provide original knowledge about disciplinary language usage, writing development, and disciplinary literacy patterns that will enhance teaching and learning resources in Hong Kong.
Project Reference No.: UGC/FDS24/H13/25
Project Title: Towards a Typology of Progressive Marking in Sinitic Languages – Synchrony and Diachrony
Principal Investigator: Dr LU Wen (PolyU SPEED)
Abstract
In the literature, a general scarcity of research on Sinitic grammars may be observed, by and large due to the long-standing misconception that there is a ‘universal Sinitic grammar’ (since Chao 1968). However, in recent years we have witnessed a substantial growth in the research on ‘dialectal’ grammar, including studies on the tense and aspectual systems of Chinese in general, such as Xiang (1997)’s work on on Liancheng Hakka, Li (1998)’s research on Suzhou Wu, Wu (1999)’s study on verbal particles in Xiang Chinese, Cai (2006)’s comparative analysis of the progressive and continuative aspect in Wu and Min Chinese, and Zhuang (2007)’s diachronic account of two aspectual marker kai 開 and li 裏 in the late 19th century, among others.
Nevertheless, methodology-wise, much research on dialectal grammar, including verbal aspect, heavily relies on linguistic elicitation based on translation tasks, namely questionnaires which are modelled on Mandarin. Sample-wise, those studies are neither balanced nor comprehensive across major Sinitic varieties, hence neglecting potential variations in the major devices for aspect marking; goal-wise, they aimed at providing a descriptive account in general, rather than an explanatory analysis of synchronic data as the result of areal convergence and language contact, and/or of a diachronic analysis due to grammaticalization from different etyma of verbs.
Despite the importance of verbal aspect in the typology of the tense, aspect and mood systems across Sinitic varieties, and for the understanding of the Sinitic grammars, works dedicated to a typological account of progressive aspect markers both from a synchronic and from a diachronic perspective are lacking. This study hence sets out to bridge this gap, by investigating a specific aspectual category, namely the progressive aspect, in a balanced sample pool of an estimation of 66 Sinitic languages, containing at least one variety for each major subgroup of the ten Sinitic groups. We will combine the Questionnaire on the Progressive Aspect (Bertinetto and Dahl n.d.), conceived for a typology of the progressive aspect in the world’s languages, and Yue-Hashimoto (1993: 80-81)'s questionnaires on progressive aspect from her Comparative Chinese Dialectal Grammar, in order to better appreciate features unique to Sinitic. In specific, we aim to:
(1) Provide a comprehensive depiction of progressive aspect marking across Sinitic varieties;
(2) Conduct a typological and areal analysis of the patterns observed in Sinitic varieties; and
(3) Explain possible source verbs and grammaticalization pathways for major types of aspectual markers.
This project will promote scientific excellence as the first comprehensive endeavour on the typology of progressive aspect marking across Sinitic languages, both in synchronic and in diachronic perspective. Furthermore, this project will generate a comprehensive archived database on the progressive aspect, facilitating the good practice of knowledge sharing among scholars and the general public. Lastly, local communities will benefit from diffusion activities organized by the project team during fieldwork to increase their language awareness and to better preserve the local culture embodied in dialects.
Project Reference No.: UGC/FDS16/E19/25
Project Title: Integrating AI-driven techniques and advanced IoT monitoring to assess seawater intrusion impacts on inland riverine systems
Principal Investigator: Dr LU Yi (HKMU)
Abstract
Seawater intrusion into rivers and freshwater systems is a growing global threat, particularly in coastal regions like Hong Kong. This phenomenon contaminates drinking water, harms aquatic life, and disrupts ecosystems. Climate change and rising sea levels are making the problem worse, with heavier rainfalls and stronger typhoons pushing seawater further inland. Hong Kong’s rivers, vital for communities and agriculture, are particularly vulnerable, yet there is limited research on how seawater intrusion affects these local waterways.
This project addresses this challenge by integrating smart technology, advanced computer modeling, and real-time monitoring. Using Internet of Things (IoT) sensors placed along rivers, we will collect data on water quality—like salt levels, sediment cloudiness, conductivity and oxygen content to track the inland movement of seawater. These sensors, combined with weather and tidal data, will feed into AI-driven models that predict when and where seawater intrusion might occur.
Building on years of research in Hong Kong’s Lai Chi Wo (LCW) catchment, the effects of storms and farming on river health have been mapped, this project introduces new tools. Drones will scan river areas to detect changes, while computer simulations replicate how tides, rainfall, and climate shifts drive seawater into rivers. By blending these approaches, we will create an early-warning system that alerts communities and policymakers to risks. classifying threats as “Emergency”, “Watch and Act” or “Advice”. For example, farmers could receive alerts to avoid irrigation during high-salinity periods, while governments could prioritize floodgates or conservation measures.
The goal is to protect Hong Kong’s freshwater resources and offer a blueprint for other coastal regions facing seawater intrusion. By transforming data into actionable insights, this project aims to safeguard ecosystems, support sustainable water use, and build resilience against climate change - ensuring clean water for future generations.
Project Reference No.: UGC/FDS16/M19/25
Project Title: Deciphering the Role and Mechanism of Rab-like Protein 1A (RBEL1A) in Cancer Cell Migration and Metastasis
Principal Investigator: Dr LUI Ki (HKMU)
Abstract
According to the World Health Organization, approximately 20 million new cancer diagnoses and 9.7 million cancer-related deaths were estimated in 2022 (the most updated figure) and the number would double by 2050. These alarming statistics highlight the inadequacy of our current understanding of cancer. Therefore, there is an urgent need for more research, particularly in the areas of cancer cell migration and metastasis, which are the primary causes of cancer-related mortality.
Rab-like protein 1A (RBEL1A) was first discovered and characterized by our team in breast cancer. Our research, along with studies from other groups, has demonstrated abnormally high RBEL1A expression in various cancers compared to their normal tissues. RBEL1A functions as a cell proliferation enhancer via activation of various cell proliferation and survival pathways, while it suppresses various cell death signals. High levels of RBEL1A in cancer are associated with lower patients’ survival rates and reduced life expectancy, which are linked to metastasis—a process where cancer spreads from the primary site to other organs, leading to over 90% of cancer-related deaths. Although both our group and others have identified this correlation, the specific mechanisms through which RBEL1A promotes cell migration and metastasis remain largely unexplored. By uncovering these precise mechanisms, we aim to pave the way for RBEL1A-targeted therapies that could help cancer patients to reduce metastasis in the future.
In this proposal, we present clinical big data from approximately 4,000 cancer patients, demonstrating that RBEL1A is frequently up-regulated in various cancers. This up-regulation correlates with poorer survival rates and increased metastasis in cancer patients shown by the big data. In vitro data further support the critical role of RBEL1A in cell migration. With elevated cell migration signals, cells have a higher metastatic potential in cancer invasion. Additionally, our preliminary data suggest that RBEL1A suppresses neprilysin (MME), a cell surface enzyme responsible for degrading the cell migration signal peptide endothelin 1 (ET-1). We will show that cancers with high RBEL1A expression inhibit MME, thereby preserving ET-1 and promoting metastasis. We hypothesize that RBEL1A suppresses MME to activate RhoA, a key regulatory GTPase that enhances cell movement by facilitating the formation of actin filaments. We will demonstrate that inhibition of RBEL1A can effectively block cell migration and metastasis by using ex vivo ET-1-mediated cell migration assays.
Furthermore, our experimental data, along with the AlphaFold AI-docking model, indicate that RBEL1A interacts with Actin-Related Protein 3 (ARP3), a crucial protein that promotes cell migration by forming migration branches. These branches facilitate the formation of additional actin filaments, which in turn generate cell protrusions that drive cell crawling movement. We hypothesize that RBEL1A associates with ARP3 to augment the formation of actin filaments. To support this, we will map the interaction regions between RBEL1A and ARP3 and demonstrate that RBEL1A enhances ARP3-mediated actin filament formation. More importantly, disrupting the RBEL1A-ARP3 interaction reduces actin filament formation and inhibits cell migration and metastasis by using the ex vivo assays mentioned above. This research project focuses on utilizing in vitro and ex vivo models to dissect the molecular roles of RBEL1A in metastasis mechanisms observed in human data. Upon completing this study, in vivo experiments will follow in the next phase of research study in the future.
In summary, this project aims to discover novel knowledge of the molecular regulation of RBEL1A in cancer cell migration, and demonstrates that inhibiting RBEL1A has therapeutic potential for the development of RBEL1A-targeted therapies in treating cancer metastasis.
Project Reference No.: UGC/FDS14/E01/25
Project Title: Robotic Mobile Fulfillment Systems: A New Item-on-Pod Replenishment Approach
Principal Investigator: Dr MA Hoi-lam (HSUHK)
Abstract
Robotic mobile fulfillment systems (RMFSs) are a variety of parts-to-pickers systems, which are widely used by many e-tailers (e.g., Amazon) and e-commerce center (e.g., DHL) nowadays in their intelligent warehouses to handle the huge variety of small products and massive orders with small item quantities from online sales. According to a Forbes reported in 2024, Amazon Prime Date sales reached USD 14.2 billion spending over a two-day period. In China, JD Logistics operates over 1,600 warehouses, handling over 1M online daily sales orders. As such, the efficiency and stability of RMFSs has become very crucial.
However, the existing shortcomings in RMFSs are its item-on-pod replenishment mechanism. A smooth item-picking process starts with having autonomous mobile robots (AMRs) moving the pods with the required items from the storage zone to the workstations for item-picking. However, prior literature commonly assumes that the availability of items on pods is unlimited, which is obviously unrealistic. Thus, in the current industrial practice as we learnt from our supporting company, pods will be sent for replenishment only when the required items are found unavailable by pickers during the item-picking. This seriously affects the RMFSs operations, causing disruption and rework. Consequently, the system efficiency and stability are seriously jeopardized.
To overcome the existing shortcomings, we propose a novel item-on-pod replenishment approach to avoid item-unavailable on pods during the item-picking and unnecessary replenishment. To achieve so, we will develop a novel customized efficient algorithm for this decision-support tool. With the support of a leading e-commerce center from Hong Kong, we will conduct numerical experiments and pilot tests by using real and semi-real data to evaluate the proposed decision-support tool and explore the importance of the proposed item-on-pod replenishment approach to RMFS efficiency. This study provides not only theoretical breakthroughs in RMFS studies but also practical contributions to industrial applications. It can enhance system efficiency and stability by avoiding disruptions due to unavailable items on pods. Meanwhile, it can increase system responsiveness to satisfy practical business needs. Our findings can provide novel insights into RMFSs operations management. This novel item-on-pod replenishment approach is not only essential to support the rapid development of e-tailers and e-commerce centers, but also very valuable for our teaching in warehouse operations and management subjects as teaching materials.
Project Reference No.: UGC/FDS16/E27/25
Project Title: Adaptive Log Analysis Framework for Proactive and Secure Software Reliability
Principal Investigator: Dr MA Xiaoxue (HKMU)
Abstract
The reliability and maintenance of software systems are essential in various industries, including finance, e-commerce, telecommunications, healthcare, and transportation, all of which rely heavily on online services and systems. These sectors require high-quality and stable software systems to ensure their operations run smoothly. Typically, run-time system information is recorded as log data, which operations and maintenance (O&M) engineers analyze to support monitoring, activity tracking, and troubleshooting. Given the substantial volume of data generated daily, automated log analysis has emerged as a prominent research focus in recent years, aimed at alleviating time and cost burdens. A critical aspect of this process involves effectively converting unstructured logs into structured formats (log parsing) to facilitate efficient analysis, alongside accurately identifying log anomalies that could lead to system failures, performance issues, or service interruptions.
Despite the promise of advanced machine learning (ML) and deep learning (DL) techniques, several key hurdles remain: 1) a disconnect between research solutions and real-world practitioner needs, 2) poor generalizability due to reliance on constrained datasets, 3) vulnerability to privacy attacks when handling sensitive logs, and 4) inadequate explainability or inaccuracy of results, which undermines trust and hinders widespread adoption. To address these challenges, we propose the Adaptive Intelligent Framework for Proactive sOftware Reliability and Maintenance (AIFORM), a modular, integrated solution designed to advance the state of log analysis. AIFORM prioritizes both performance and security while offering practitioners actionable insights. The framework's key innovations include: (1) data diversity and augmentation, achieved through adversarial data generation and validation to produce high-quality, diverse log datasets that enhance model generalizability and robustness; (2) reliable security measures, which involve assessing vulnerabilities to privacy attacks and implementing novel defense mechanisms to safeguard sensitive log data; (3) explainable anomaly detection, leveraging large language models (LLMs) in conjunction with explainable AI techniques to provide interpretable and accurate anomaly detection results; (4) adaptive log parsing, integrating heuristic rules, statistical and semantic information, and utilizing reasoning process to automatically parse data; and (5) a practitioner-centric design, explored through in-depth studies to align the framework with industry needs and expectations.
AIFORM’s modular design ensures adaptability across diverse systems and industries, offering a scalable and explainable solution for comprehensive log analysis. By addressing the limitations of current methods, AIFORM will advance software reliability and maintenance, supporting critical operations across sectors and driving innovations in automated log analysis.
Project Reference No.: UGC/FDS25/M01/25
Project Title: Design and pharmacological characterizations of multifunctional anti-Parkinson’s drug leads via regulating neuroinflammation in vitro and in vivo
Principal Investigator: Dr MAK Shing-hung (THEi)
Abstract
Neurodegenerative disorders, such as Alzheimer’s disease (AD) and Parkinson's disease (PD), have been regarded as one of the most significant challenges in modern medicine due to the uncertain pathogenesis. Multiple pathological factors, including environmental and genetic, have been suspected to cause PD. Clinical available single-targeted drugs only offer limited therapeutic benefits to PD patients. Progressive loss of dopaminergic neurons in substantia nigra has been closely associated with the development of PD. Thus, “Multi-target Directed Ligands (MTLDs)” strategy have been proposed to develop the novel candidates for the treatment of PD. Moreover, sustained neuroinflammation might play a vital role in PD by promoting the neural degeneration.
Recent studies have shown that α-synuclein aggregation regulates the key components, such as NLR family pyrin domain containing 3 (NLRP3) and nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB), in the signaling pathway of neuroinflammation. Thus, targeting the pro-inflammatory signaling targets such as NLRP3 and / or NF-κB might be the effective strategy to treat PD. By using a sequential combination of ligand and structure-based virtual screening techniques, as well as molecular docking analysis, we optimized and designed several novel virtual drug leads, which aim on the regulation of neuro-inflammatory response by targeting the key inflammatory signaling molecules including NF-κB and NLRP3. Thus far, we have successfully synthesized several novel drug leads. The preliminary results shown that these lead molecules do not possess cytotoxicity in PC12 cells.
In this proposal, we will begin with the synthesis of novel anti-PD MTDLs and probe the inhibitory effects on different neuroflammatory targets. Enzymatic assays, ELISA and Western blot assays will be employed to evaluate the abilities of the drug leads on variety of molecular and/or cellular targets, such as expression and the aggregation of misfolded α-synuclein, and the potential therapeutic targets including NF-κB and NLRP3. Then, the drug leads shown significant inhibitory effects will be selected to assess the effects on neuroprotection and/or promotion of abnormal activation of microglial cells in vitro. Furthermore, the novel MTDLs’ impact on neuroinflammatory responses will be investigated by analyzing the modulation of pro-inflammatory cytokines release. Particularly, the key components regulating NLRP3 inflammasomes mediated neuroinflammatory pathway will be focused.
In conclusion, this study aims to develop a series of multifunctional drug candidates mainly targeting neuroinflammatory response. By utilizing a comprehensive approach, the underlying mechanisms will also be investigated. The outcomes will contribute to the development of a potential therapeutic intervention, addressing the urgent needs in this challenging field.
Project Reference No.: UGC/FDS16/M13/25
Project Title: Prolonged flooding and elevated salinity induced by climate change: response of mangrove sediment microbiome from salt pan
Principal Investigator: Dr MO Wing-yin (HKMU)
Abstract
Mangroves are vital coastal forests that protect shorelines, support a diverse array of marine life, and play a critical role in mitigating climate change through carbon sequestration. At the heart of these ecosystems are the sediment microbiomes—complex communities of microorganisms that drive nutrient cycling, facilitate organic matter decomposition, and regulate greenhouse gas emissions. However, these microbial communities are extremely sensitive to changes in their surrounding environment. Even subtle shifts in hydrology, oxygen availability, and salinity can disrupt their delicate balance. With climate change accelerating sea-level rise and increasing saltwater intrusion, the natural conditions that sustain these sediment microbiomes are being rapidly altered, placing the entire mangrove ecosystem under stress.
In addition to the impacts of climate change, many mangrove areas have been converted into salt pans and fish farms, dramatically modifying their natural conditions. These human-modified sites, which exhibit controlled flooding regimes and naturally high salt levels, now mimic the projected environmental scenarios of the future. Despite these significant changes, no study has yet experimentally examined how the combined effects of elevated salinity and prolonged flooding influence the structure and function of sediment microbiomes during mangrove restoration. Abandoned salt pans, therefore, offer a unique natural laboratory to assess the resilience of these microbial communities and the overall efficacy of different restoration methods under future climate conditions.
This proposed project is structured into two parts designed to address these critical knowledge gaps. In Part 1, we will conduct extensive field investigations across three distinct types of sites: (1) mangrove areas restored in salt pan through manual replanting, (2) mangroves have naturally colonized abandoned salt pans, and (3) natural mangrove forests that serve as reference sites. These sites provide an ideal setting for comparing the establishment and functional performance of sediment microbiomes across different restoration strategies. We will assess both the microbial community composition and the spatial distribution of functional genes that underpin key biogeochemical processes, including the regulation of greenhouse gas emissions.
In Part 2, we will select the sites that exhibit the highest and lowest greenhouse gas emissions from our field survey for more investigations. Controlled experiments will be designed to simulate future environmental conditions, specifically extending flooding durations and further increasing salinity levels. By monitoring shifts in microbial metabolic pathways and tracking changes in the abundance and diversity of functional genes, we aim to determine which restoration approach offers greater resilience under stress while effectively moderating greenhouse gas emissions.
This phase will provide a mechanistic understanding of how sediment microbiomes respond to, and potentially mitigate, the adverse effects of environmental stressors. The findings from this research will deliver critical insights into the belowground processes that support the ecological functions of mangroves. By clarifying the interactions between sediment microbiomes, environmental stress factors, and greenhouse gas fluxes at a mechanistic level, our study will provide a robust scientific foundation for improving mangrove restoration practices. These insights will be vital for conservation managers, ecologists, and policymakers tasked with the development of climate-resilient restoration strategies that not only revive these essential ecosystems but also enhance their capacity for carbon sequestration and reduce harmful greenhouse gas emissions. Ultimately, our research will contribute to safeguarding the long-term health and functionality of mangrove forests in an era of rapid climate change.
Project Reference No.: UGC/FDS15/H17/25
Project Title: Socialist Realism and Japanese Reportage Painting during the Cold War
Principal Investigator: Dr NG Camellia Ni-na (Shue Yan)
Abstract
The Cold War, often perceived as a concluded historical chapter, continues to demand scholarly exploration of its enduring cultural legacies. While existing scholarship frames the mid-to-late 20th century through the ideological dichotomy of the US–UK-led capitalist bloc and the Soviet–China-aligned communist sphere (Christopher 2014; Davies 2022; Westad 2017), contemporary Japanese art history remains narrowly focused on post-war anticommunist narratives tied to Japan’s role in US containment strategies (Broinowski 2016; Kitazawa, Kuresawa, and Mitsuda 2023; Mikkonen, Scott-Smith, and Parkkinen 2019). This emphasis has overshadowed critical geopolitical intersections, particularly the influence of Asia-Pacific dynamics on Japan’s artistic evolution. Notably, avant-garde movements such as Gutai and Mono-ha dominate academic discourse, yet the politically charged Reportage painting movement of the 1950s, a socialist realist response to US militarization and Japan’s perceived subordination, reveals deeper transnational tensions warranting scrutiny (Carroll 2024).
Emerging in the early 1950s, Reportage painting combined the approachable aesthetics of socialist realism with pointed critiques of U.S. military influence and Japan’s contested sovereignty (Jesty 2018). The movement’s rise coincided with the Korean War (1950–1953), a pivotal conflict that saw China’s defiance of U.S.-led UN forces redefine East Asian power structures. Following the 1953 Panmunjom Truce—interpreted by many as a U.S. concession— Japanese intellectuals reassessed Cold War alliances, revitalizing socialist thought in public discourse (Kami 1996; Namiko 2017). Although scholars have analyzed Reportage’s role in domestic socio-political dissent (Merewether and Hiro 2007; Western 2004), its transnational aspects remain overlooked. These include its stylistic departure from Soviet and Chinese socialist realism, as well as its dialogue with Hong Kong, then a burgeoning hub of global exchange. Further study is needed to unravel how Japanese artists reinterpreted socialist aesthetics within their distinct ideological “field,” positioned at the intersection of capitalist and communist blocs.
Employing Pierre Bourdieu’s field theory, this study analyses how reportage practitioners navigated political ideologies through institutional engagements, such as municipal bodies, arts funding systems, exhibitions, and media. It argues that reportage, rooted in Russian influences but catalyzed by the Korean War, reflects an organic evolution of Japan’s intellectual–artistic frameworks rather than a mere derivative of socialist theory. By situating the movement within broader Asia-Pacific Cold War geopolitics, including Hong Kong’s role in disseminating ideological currents, this research challenges reductive East–West binaries and illuminates Japan’s complex positioning as both a US ally and a site of socialist artistic innovation.
Project Reference No.: UGC/FDS14/P03/25
Project Title: Optimal sub-sampling strategies for tensor factor analysis of time series data
Principal Investigator: Dr NG Chi-tim (HSUHK)
Abstract
Modern technology allows researchers to collect tensor-valued data from different time points. For example, the keyword search frequencies of a search engine about 𝑃1 = 100 health related keywords in 𝑃2 = 100 regions of a country over a period of 𝑇 = 52 weeks. Statistical analysis on such a matrix-valued time series data faces difficulty related to the curse of dimensionality.
Consider a general situation of tensor-valued time series data under certain tensor factor analysis models. The goal of this research is to propose a sampling strategy to select an affordable subsample from the full data so that the time-dependence structure of the data and all factor loadings can be estimated with a certain degree of accuracy. A two-stage approach is developed for such a purpose. First, a preliminary estimation is obtained based on a random subsample from the full time series. Second, a genetic algorithm-based method is proposed to obtain an optimal sampling strategy from the preliminary estimation for optimizing the mean squared estimation error. In addition, an objective function is introduced to describe the balance property related to the occurrence frequencies of each component/dimension in the sample. To the best of our knowledge, this is the first research to consider the sampling issues on the tensor-valued time series data analysis.
Computational methods and software are developed for the proposed sampling and estimation strategies. Statistical theory is established to explain the relationship between the accuracy of the estimation and the choice of the sample size.
The research results are important for the tensor-valued data analysis in engineering and many academic disciplines in the future. Empirical examples like keyword trend analysis will be used to illustrate the proposed sampling strategy and estimation methods. The information about the keyword-effects, geographical-effects, and the time-effects on the keyword search trend can therefore be extracted within an affordable computational time. The time evolutionary process of the underlying driving forces that govern the development of all components/dimensions can be studied simultaneously. The proposed method can also be applied to data involving interaction between countries, for example, the exports, transportation cost, etc. between the departure country and the destination country of different items over time.
Project Reference No.: UGC/FDS24/E27/25
Project Title: Golden Bamboo Grass (GBG)-Reinforced Polylactic Acid (PLA) Biocomposite for Sustainable Toys
Principal Investigator: Dr NG Sun-pui (PolyU SPEED)
Abstract
The global production of plastics worldwide has witnessed a significant increase over the past 70 years, with a staggering 400 million metric tons produced in 2022 alone. If plastic waste is not effectively managed through recycling, incineration or landfilling, plastic pollution will become an inevitable environmental crisis for humanity. Among different plastic products, toys contribute significantly to this pollution issue. For example, the average annual production of LEGO® building toys is around 36 billion pieces, and over 600 billion plastic parts have been made since 2015. Acrylonitrile butadiene styrene (ABS) is the conventional material used in manufacturing rigid toys, and it is known for its durability and longevity. For instance, 80% of LEGO® building bricks are made from ABS. However, ABS is not biodegradable, and it takes 1,300 years for ABS to break down in the oceans, according to a report from Yale School of the Environment. Nevertheless, both mechanical and chemical recycling methods are not suitable for small ABS toy pieces like LEGO® bricks. The bricks’ small sizes and shapes make them liable to get stuck in standard recycling machinery or slip through sorting screens. Moreover, while chemical recycling is theoretically sustainable, it is costly, energy-intensive, and environmentally burdensome in practice. To address this issue, the LEGO® Group commenced research on the use of recycled polyethylene terephthalate (PET) from drink bottles in 2021. However, all relevant research efforts were abandoned in September 2023, as the required recycling processes resulted in higher carbon dioxide emissions and manufacturing costs. Therefore, a viable research direction is to employ bioplastic reinforced by biodegradable lignocellulosic fillers or fibres in order to possess mechanical strength comparable to ABS plastic.
In recent years, a novel crop known as Golden Bamboo Grass (GBG) with an exceptionally high cellulose content (up to 75.04%) was produced in Guangxi Province of China, which was cultivated from modified types of Arundo donax L. GBG is a rhizomatous perennial grass that grows rapidly and reproduces vegetatively. It is regarded as one of the popular crops for biomass production on marginal and degraded lands under a variety of unfavorable conditions, including drought, salt, waterlogging, high and low temperatures, and heavy metal stress. Due to its adaptive and resilient growing behavior, the harvest cycle and year are 3 months and 20 years respectively and it becomes an accessible natural material with abundant availability. Its potential application in bioplastics is noteworthy due to its carbon-neutral properties, chemical content and rapid growth rate in various tough environments and cultures. Therefore, this project proposes to incorporate GBG as the reinforcement fibre in polylactic acid (PLA) bioplastic to make fully biodegradable toys. However, the hydroxyl (-OH) groups on GBG cellulosic fibers make them hydrophilic, whereas PLA has non-polar methyl groups (-CH3) that make them hydrophobic. Such incompatibility causes weak adhesion between GBG fibre and PLA, resulting in decreased mechanical strength for the entire composite material. This project aims to address these issues by developing a GBG-reinforced PLA biocomposite material with improved mechanical properties through the use of fibre surface treatment and montmorillonite (MMT) nanoclay as filler reinforcement. The optimal contents of GBG fibre and MMT nanoclay in PLA will be determined from their morphological, mechanical and thermal properties. Furthermore, a quantitative study of soil biodegradability will also be conducted to determine the degradation rate of the GBG-reinforced PLA biocomposite.
Project Reference No.: UGC/FDS16/M02/25
Project Title: Mechanisms of Cadmium Dynamic Migration Mediated by Root Iron Plaques in Different Mangrove Species
Principal Investigator: Dr PAN Min (HKMU)
Abstract
Unsustainable industrial and agricultural practices have resulted in the introduction of heavy metals into the environment, which pose serious threats to human health and marine ecosystems. Cadmium (Cd), a widely distributed heavy metal with high toxicity, persistence, bioaccumulation, and the potential to disrupt biological processes, is of particular concern in marine ecosystems. Mangroves, which are vital woody plants in the intertidal zones of tropical and subtropical regions, have significant ecological value. They are often referred to as "coastal guardians", function as barriers to tidal/wave surges caused by tropical cyclones, and play crucial roles in environmental purification and pollution control, as they frequently face different metal stresses in the marine ecosystem. Remarkably, they have developed mechanisms to defend themselves and resist pollutants. For example, iron plaques (IPs) on the surface of mangrove roots can act as barriers, effectively fixing and transforming heavy metals in rhizosphere sediments. This process strongly influences the migration and transformation of heavy metals in wetlands. The formation of root IPs is not a steady process, as it can be affected by the iron status in the sediment, root radial oxygen, and mangrove plant species. It is vital to understand the dynamic formation processes of root IPs in different mangrove species and to determine how root IPs contribute to the responses and resistance of mangroves to Cd contamination. Do root IPs block Cd accumulation in mangroves, and do they also affect nutrient uptake by the plants? Research on the mechanisms by which root IPs in mangrove wetlands affect dynamic Cd migration, nutrient transport and Cd transporter genes in mangrove roots is still in its infancy. Conventional detection methods cannot continuously monitor dynamic Cd migration processes in plants. The newly developed noninvasive microtest technique (NMT) has potential, but previous applications of this technique have focused mainly on terrestrial plants, and NMT is underexplored in mangrove studies.
Therefore, we aim to conduct in-depth and systematic studies on the temporal formation processes of mangrove root IPs in different mangroves and explore their roles in resistance to Cd contamination by the mangroves and their effects on nutrient transportation in marine ecosystems. The objectives of this research are as follows: (1) to evaluate the temporal formation mechanisms of root IPs in the four widely distributed mangrove species with low and high concentrations of exogenous Fe2+ in the sediment; (2) to compare the influence of root IPs on the stress response mechanisms to Cd-contaminated sediments among the four species, including the dynamic migration of Cd from rhizosphere sediment to root IPs, Cd uptake and Cd bioaccumulation in different plant tissues; (3) to assess the effects of root IPs on the dynamic migration, bioaccumulation and transportation of nutrients, including K+, H+, Na+, and NH4+, from rhizosphere sediment to root IPs and then to different tissues in the presence of different Cd contaminations; and (4) to explore the roles of Cd transporter genes in the root IPs of different mangrove plant species exposed to different levels of Cd contamination by a transcriptomic approach. The results will provide valuable insights into root IP formation and resistance to Cd contamination in widely distributed mangrove species, highlighting the adaptive mechanisms by which mangroves thrive in challenging environments. In this research, NMT will be utilized to assess the dynamic migration of Cd and nutrients in specific mangrove species. Understanding the species-specific dynamics of Cd migration and accumulation from the rhizosphere to mangrove roots with IPs is crucial for applying mangrove resources to enhance ecosystem resistance to heavy metal pollution in wetlands.
Project Reference No.: UGC/FDS15/H31/25
Project Title: “Women’s Work for Women”: The New Intellectual Woman in the Republic of China, Wu Yi-fang (1893-1985)
Principal Investigator: Dr PANG Suk-man (Shue Yan)
Abstract
This research project examines the career of Dr. Wu Yi-fang (1893-1985) as president of Ginling College, a Christian higher education institution based in Nanjing. This research project focuses on two aspects of her personal life and work experiences to gain a deeper understanding of the long 20th century, during which China underwent momentous changes.
First, a case study of Dr. Wu’s professional career allows for a close investigation of the rise and fall of Christian universities and colleges during the tumultuous years of Republican China (1912-1949). In the late 1920s, the Nationalist government proceeded to secularize and Sinify Christian universities, paving the way for Wu Yi-fang’s ascendance as the first and only Chinese president of Ginling College. Dr. Wu reconciled the two seemingly competing threads of educational ideas: the Christian and the patriotic. Dr. Wu herself was a political activist whose political career culminated in 1945 when she represented China in signing the Charter of the United Nations. This proposed research project will significantly reflect the interconnectedness between Christianity and patriotism.
Second, the project highlights Dr. Wu as a pioneer of modern Chinese womanhood. She was an exemplar of a new breed of intellectual Chinese women in the first half of the 20th century. Her womanhood resulted from the newfound educational opportunities for women, along with career prospects, widely circulated and accepted feminist ideas, and the influence of Christianity. As a woman who remained single throughout her life, Dr. Wu’s womanhood was heavily tied to her career as a leading educator and political activist.
Project Reference No.: UGC/FDS13/E02/25
Project Title: IoT-Based Artificial Intelligence Maturity Model for Predicting the Development of Early Strength in Concrete
Principal Investigator: Dr PENG Yixin (Chu Hai)
Abstract
Determining the early-age compressive strength of concrete is crucial for quality assurance and structural safety in construction projects. While the maturity method provides a means to non-destructively assess concrete strength development based on temperature history, current practices have limitations. Establishing unique maturity curves through extensive laboratory testing for each individual concrete mix hinders the practical application of real-time strength prediction on construction sites. In addition, conventional maturity methods rely on a certain “maturity” determined by the accumulated effect of temperature may not accurately reflect the varying temperature conditions encountered in field placements.
This study aims to address these limitations by developing an artificial intelligence (AI) maturity model for real-time prediction of concrete compressive strength. The model will be trained to forecast concrete strength at different early ages based on measurable parameters like mix design, curing temperature, and humidity, without the need for individual strength-maturity relationships.
An Internet of Things (IoT)-enabled monitoring system will be implemented to automatically collect temperature and humidity data from in-situ concretes with sensors. Wireless transmission using network technologies will send the sensor readings to a laboratory curing chamber for parallel testing of companion concrete specimens. This approach facilitates developing an AI model that considers the time-varying curing conditions closer to field scenarios.
The AI model will be trained using a comprehensive dataset compiled from past research results as well as site-specific data collected locally. While these data can supply information on different mix designs and curing scenarios, the on-site measurements will provide variable temperature histories. Both the historical data and field readings will be used jointly to establish the predictive AI formulation. The IoT experimental data, though limited in volume, would serve to incorporate time-varying curing conditions not represented in prior fixed-temperature test data. Machine learning techniques will be applied to develop an AI maturity model relating measured data and properties. Model validation will use concrete mixes independently of training data. Case studies will assess the automated real-time monitoring capability. Ultimately, this novel methodology integrating IoT sensors, wireless communications and AI techniques aims to advance the maturity method through real-time prediction of early-age concrete strength under actual fluctuating curing conditions encountered in construction.
Successful implementation of an AI maturity model could substantially promote the use of maturity method in construction through automation. This has the potential to reduce project costs from optimized construction scheduling based on real-time strength forecasting. It may also decrease embodied carbon emissions through reduced requirements for on-site concrete cube casting and testing. Overall, the proposed research aims to advance construction safety and quality management through an innovative IoT-enabled artificial intelligence approach.
Project Reference No.: UGC/FDS24/E19/25
Project Title: Thin, Soft, Sensitivity-Tunable, Low-Voltage-Driven OECT-based bioinspired E-Skin with 3D Mesostructured for tactile perception in Humanoid Robots
Principal Investigator: Dr SHI Rui (PolyU SPEED)
Abstract
This proposal aims to explore the development of a novel 3D electric skin architecture utilizing organic electrochemical transistors (OECTs) to simulate human skin tactile sensations. Unlike previous studies focused on 2D laminated skin structures, this research will leverage a 3D architecture to enhance the decoupling of pressure and stress forces, improving the sensitivity and responsiveness of the electric skin. The unique characteristics of OECTs, including adjustable gate voltages and the use of solid-state ionic liquid electrolytes, will be employed to optimize performance. This research has significant implications for advancements in prosthetics, robotics, and human-computer interaction.
Project Reference No.: UGC/FDS11/E13/25
Project Title: Multi-hybrid Video Coding via Strategic Framework with Deep Learning and Super-Resolution Context
Principal Investigator: Prof SIU Wan-chi (SFU)
Abstract
No doubt video coding is an important research topic, even though the advancement of computer technology and communication techniques can alleviate the problem. Using one hard disk is always better than using two hard disks for storage, and waiting for one minute is always better than waiting for two minutes for transmission. Conventional approach, like the Hybrid Video Coding, seems coming to the physical limit. For further break though, we have to think out of the box. Our approach on using the best state-of-the-art codec, deep learning with generative power and super-resolution techniques for building a novel video coding base-line should be a good starting point.
The overall objective of this research is to propose a novel baseline video coding kernel, which starts from down-sizing the original video frames by simple means into a smaller size. This is followed by using any currently existing video codec to code the downsized video for storage or transmission. The original video is recovered from the decoder of the video codec with super-resolution techniques by making use of the generative power of deep learning. Let us summarize in point form the objectives of this proposal.
1. We propose a new hierarchical baseline structure for video coding, which makes use of the best “off-the-shelf” video codec as the centre component, and the generative power of deep learning and the super-resolution power in deep learning as the overall structure.
2. A video codec has to be chosen to work with the baseline model and the super-resolution unit. A suitable one is not necessary to be the most recently (such as H.266, which is complicated, or even H.265 which might get very complicated licensing problem) available codec. We have to choose/design the one which allows fast decoding and allow to have high quality output.
3. We expect to propose a Reference Based Super-Resolution unit for this project which can work well with a down-sizing filter. It should be fast, adaptive to the current scheme, and hopefully allow super-resolution in steps hence progressive coding is possible.
4. For making use of the Reference Based Super-Resolution unit, a small number of frames are used as reference frames (R-frames) while other are low quality (L-frames) video frames. The study is to investigate a scheme with the right mix of R- and L-frames etc. for efficient super-resolution. This should also include the way to make down-sampling of the original video frames, which could affect the overall scheme.
Project Reference No.: UGC/FDS11/H03/25
Project Title: Loneliness and Health Literacy in older adults: A sequential mixed methods study
Principal Investigator: Prof SMITH Graeme Drummond (SFU)
Abstract
Loneliness is a highly subjective experience of deficit between the actual and desired levels of social contact it is estimated to affect around 35% of Hong Kong older adults. Loneliness increases all-cause mortality by a quarter and is associated with increased health care expenditure. Societal measures that support equity of social participation have been suggested to reduce loneliness. Globally, low health literacy (HL) is a significant health equity issue as HL is a stronger predictor of health outcomes than race, and educational attainment. Health literacy is the cognitive and social skills to access, understand, appraise, and use information to promote and maintain one’s health. Internationally, 29% to 75% of older adults have low levels of HL. Older adults’ demographic information (e.g., age, gender, marital status, education attainment, income, living arrangements), health status, and social participation are consistently related to loneliness and HL. Given low HL is a global health equity challenge for older adults, conclusive empirical evidence on its relationship with loneliness, as a public health threat, is urgently required.
In this sequential mixed method study (QUANT-QUAL), we aim to examine the relationships between loneliness and HL in community-based older adults. We hypothesize that HL is inversely associated with loneliness.
The main objectives of this proposed study are:
1. To examine the associations between HL and loneliness by controlling other known individual characteristics, physical functioning, cognitive functioning, and social participation in community-dwelling older adults attending Non-Governmental Organization social care centres in Hong Kong.
2. To describe the challenges faced by older adults with loneliness at various levels of HL.
3. To explore the meanings of loneliness through gathering narratives from older adults with various levels of HL; and
4. To develop a culturally specific tailor-made HL intervention for older people.
This study will target community dwelling older adults (aged 60 or above). In order to enhance the clinical relevance or interpretation of the association between loneliness and HL, the two variables will be analyzed as dichotomous variables. Employing a mixed methods study will allow us to quantify loneliness and HL, while those subjective feelings (e.g., shame) that may arise from low HL can be captured by qualitative methods. In phase one, quantitative survey data collection will include the use of a battery of valid and reliable measurement scales, including the 6-item Chinese version of the De Jong Gierveld Loneliness scale; Chinese version of 12 item of Short-Form Health Literacy Questionnaire (HLS SF-12), Chinese version of six items Lubben Social Network Scale, Chinese version of the 9-item Lawton Instrumental Activities of Daily Living Scale and a demographic form. In phase two, qualitative data will be collected by audio recorded individual face-to-face interview. In phase three, the findings of the quantitative and qualitative phases of the study.
Addressing this knowledge gap is merited to facilitate the provision of optimal care to older people in Hong Kong, enhancing healthy ageing.
Project Reference No.: UGC/FDS14/H01/25
Project Title: From Past to Present: A Longitudinal Study on the Evolution of Business Journalism Practice and Education in Hong Kong (2015-2025)
Principal Investigator: Dr SONG Zhaoxun (HSUHK)
Abstract
Over the past decade, Hong Kong’s business journalism landscape has undergone profound transformation, primarily driven by technological advancements, particularly the rise of artificial intelligence (AI). The proposed research project, "From Past to Present: A Longitudinal Study on the Evolution of Business Journalism Practice and Education in Hong Kong (2015-2025)," serves as the crucial second phase of my initial RGC grant project in 2014, "From the Newsroom to the Classroom: Bridging the Gap between Business Journalism Practice and Education in Hong Kong." This research aims to explore the evolution of business journalism practices and education in Hong Kong from 2015 to 2025, focusing on the integration of AI, the adoption of data journalism, and the challenges and opportunities these changes present for both newsrooms and educational institutions. By doing so, the study seeks to address pressing questions about the current state of business journalism while meticulously documenting its historical development in Hong Kong.
The media landscape in Hong Kong increasingly utilizes AI technologies. For instance, the AI Lab at Radio Television Hong Kong (RTHK) exemplifies AI's transformative impact on media production, using tools for language translation, archival footage restoration, image and video enhancement, noise reduction, and text-to-image/video generation. These innovations improve production efficiency and visual quality, indicating a shift toward AI-driven journalism. Similarly, The South China Morning Post (SCMP) automates financial news generation, reflecting the growing reliance on technology in business journalism.
The rapid adoption of AI in newsrooms has exposed a widening gap between industry practices and journalism education in Hong Kong. Our 2014-2015 study identified essential skills for business journalists, but AI was not a focus at that time. By 2025, AI has become central to newsroom discussions, revealing a lag in educational curricula. This research aims to tackle critical questions: How has the newsroom evolved over the past decade? What competencies do business journalists require today? How significant is the gap between newsroom practices and classroom education? And how can curricula be enhanced to effectively bridge this divide?
By systematically documenting and analyzing these developments, this longitudinal study will provide valuable insights into the practice-education gap and inform future curriculum design. The project aims to foster closer collaboration between newsrooms and educational institutions, ensuring that business journalism education remains relevant and impactful in Hong Kong. To enrich The Hong Kong Business Journalism History Database (HKBJHD), research findings will be uploaded to the website, enhancing accessibility and disseminating valuable information to researchers, students, and the public. Ultimately, these findings will empower educators and practitioners to navigate the challenges and opportunities of an increasingly AI-driven media landscape, keeping business journalism vibrant in Hong Kong.
Project Reference No.: UGC/FDS16/H24/25
Project Title: Ethical Use of AI in Education: A Framework to Evaluate Teachers' Understanding and Pedagogical Recommendations
Principal Investigator: Dr TANG Ko-wai (HKMU)
Abstract
The widespread adoption of Artificial Intelligence (AI) technologies in education has sparked an urgent discussion about the ethical implications of their use. Despite significant advancements in AI capabilities, a considerable gap remains in understanding how educators perceive and address AI ethics within educational settings. This study aims to bridge this gap by exploring teachers' awareness, practices, and needs regarding AI ethics in Hong Kong, using the Technological Pedagogical Content Knowledge (TPACK) framework and UNESCO’s AI Competency Framework for Teachers (AI CFT).
The primary research objectives of this study encompass two dimensions: first, evaluating the current understanding among school teachers of fundamental ethical issues related to AI, such as data privacy, algorithmic bias, accountability, and human agency; and second, assessing and proposing effective pedagogical strategies for teaching AI ethics to students. Focusing on these areas, this study seeks to empower educators to navigate the complex ethical landscape of AI in their teaching practices.
This study will employ a mixed-methods research design to explore teachers’ awareness, practices, and professional needs related to AI ethics. The research begins with a literature review and qualitative interviews to identify the key dimensions of AI ethics such as data privacy, bias, accountability, and human control. The insights gained from these interviews will inform the development of a structured survey and scenario-based evaluations to assess teachers' competencies and moral reasoning regarding AI ethics in educational contexts. These instruments will be content-validated by experts before conducting a cross-sectional survey with a diverse sample of teachers in Hong Kong.
The expected outcomes of this study include a comprehensive assessment of teachers’ current knowledge and practices regarding AI ethics, as well as a thorough identification of best practices and areas requiring further development. The findings will inform actionable recommendations aimed at enhancing teacher-training programs, fostering ethical awareness, and promoting responsible use of AI in educational settings. Finally, this study contributes to the broader discourse on AI ethics in education by providing valuable insights to educators, policymakers, and AI technology designers.
Project Reference No.: UGC/FDS16/M22/25
Project Title: Deciphering the pathophysiological role of platelet PANoptosis in sepsis
Principal Investigator: Prof TANG Wai-ho (HKMU)
Abstract
Sepsis is a life-threatening organ failure caused by a dysregulated body response to infection, and it has become a globally recognized public health issue with leading incidence and mortality rates. The differentiation between conditions like disseminated intravascular coagulation (DIC) and thrombotic microangiopathy (TMA) in sepsis is challenging and crucial, as it can significantly impact patient management and outcomes. Unfortunately, current clinical treatment of sepsis remains largely empirical antibiotic therapy and symptomatic support. The “gold standard” for diagnosis, blood culture is time-consuming thus may delay the accurate treatment of sepsis. This underscores the urgent need for efficient biomarkers to monitor the progression of sepsis. Platelets have been found to contribute significantly to both hemostasis and immune functions in sepsis. However, the role of platelets and the regulatory mechanisms involved in sepsis-induced thrombosis, excessive inflammation, and thrombocytopenia remain unclear.
Others and our group have demonstrated that platelets undergo pyroptosis in sepsis, exacerbating inflammation. PANoptosis is an inflammatory programmed cell death regulated by the PANoptosome complex, characterized by pyroptosis, apoptosis, and/or necroptosis. Our preliminary experiments showed that pyroptosis, apoptosis, and necroptosis ensued simultaneously in the platelets of patients with sepsis, suggesting the presence of PANoptosis. The components of PANoptosome (Pyrin and AIM2) were found to be highly expressed and co-localized. This leads to the scientific hypothesis that PANoptosis acts as a molecular switch, controlling pyroptosis, apoptosis, and necroptosis in platelets, and that different groups of platelets may promote thrombosis, excessive inflammation, and reduction of platelet (i.e. thrombocytopenia) during sepsis.
With the use of cell culture studies and animal models, the proposed project aims (i) To investigate the presence of platelet PANoptosis in sepsis and identify the components of PANoptosome in platelets; (ii) To identify the activators of platelet PANoptosis to provide a new theoretical basis for exploring the pathogenesis and treatment of sepsis; and (iii) To assess the application of targeting PANoptosis-related components to develop novel biomarkers for sepsis.
Specifically, in task (i), a sepsis mouse model will be constructed, the expression of PANoptosis-related proteins in sepsis platelets and the expression in the platelets of control mice induced to cell death in vitro (pyroptosis, apoptosis and necroptosis), will be evaluated, as well as the expression of PANoptosome-related components in the sepsis platelets. In task (ii), the expression of inflammatory and immune response-related factors during sepsis, i.e. the activators of platelet PANoptosis, will be explored in a sepsis mouse model, then neutralizing antibodies against these activators will be injected into sepsis model mice to observe if PANoptosis can be inhibited. In task (iii), platelet-specific gene knockout mice will be constructed, and the key signaling components in the process of PANoptosis in sepsis will be inhibited or neutralized.
Our findings would provide new theoretical insights for understanding the pathogenesis of sepsis and identifying potential biomarkers and therapeutic targets. The exploration of platelet PANoptosis in sepsis may also offer better understanding of the pathogenesis of sepsis-induced thrombocytopenia.
Project Reference No.: UGC/FDS16/E09/25
Project Title: Multi-chain Cryptocurrency Fraud Detection Model Based on Transformer with Graph Representation
Principal Investigator: Dr TAO Bishenghui (HKMU)
Abstract
In an era of rapid digitalization, the proliferation of cryptocurrencies has introduced both unprecedented opportunities and significant risks. Malicious activities such as telecom fraud and cryptocurrency scams threaten the security of property, resulting in staggering global losses that have reached billions of dollars. In this context, Hong Kong stands at a pivotal crossroads as a leading financial hub in Asia, endowed with a legal framework that supports cryptocurrency transactions while bearing the crucial responsibility to combat these escalating threats effectively.
This project aims to develop pioneering methodologies aimed at enhancing the detection and prevention of cryptocurrency fraud by addressing critical limitations within current practices. Specifically, we will focus on overcoming issues related to inadequate multi-chain data integration, challenges in accurately representing transaction graphs, and underperformance in existing fraud detection models. Our objectives are threefold: (1) to construct an all-encompassing Blockchain Cryptocurrency Network (BCN) database that integrates diverse multi-chain transaction data; (2) to design an advanced graph representation learning model capable of extracting intricate features from complex transactional relationships; and (3) to develop an efficient Transformer-based detection model adept at adapting across various datasets.
Our methodology is structured around three strategic tasks. First, we will undertake extensive dataset acquisition through innovative integration techniques designed to capture transactional relationships across multiple cryptocurrencies effectively. Next, we will deploy cutting-edge graph neural network approaches for extracting meaningful embeddings that reflect user behaviors within these interconnected networks. Finally, robust transformer-based detection models will be constructed for real-time identification of fraudulent activities while ensuring scalability and adaptability.
The expected outputs include the development of Graph Representation Learning and Transformer models suitable for fraud detection, academic results will be published in journals, and participation in seminars at international conferences; not limited to academic contributions, but also of great benefit to the stakeholders in the cryptocurrency ecosystem. We envision the creation of a user-friendly platform equipped with essential risk information about malicious nodes, while providing users with the necessary tools to enable them to make informed decisions around their transactions. Additionally, the initiative aims to create educational opportunities through a blockchain anti-fraud course tailored for undergraduate students, while fostering collaborative research efforts to enhance knowledge dissemination across academia and industry. By leveraging cutting-edge technologies to mitigate cryptocurrency-related risks while promoting responsible investment practices among users and researchers alike, this project not only addresses immediate concerns, but also reinforces Hong Kong's position as a secure global financial centre committed to innovation amidst emerging challenges.
Project Reference No.: UGC/FDS24/H09/25
Project Title: Impact of Customers’ Knowledge and Primary Healthcare Initiatives on Customers’ Experiences of Service Quality Attributes, Consumption Values, Place Attachment, and Behavioural Intention: Multilevel Analysis of District Health Centres (DHCs)
Principal Investigator: Dr TAVITIYAMAN Pimtong (PolyU SPEED)
Abstract
Enhancing individuals' healthy lifestyles and well-being is of paramount importance. While primary healthcare systems strive to explore various health treatments, services, and resource allocations to meet individuals' needs, significant challenges persist. These challenges necessitate further investigation into the effectiveness and sustainability of primary healthcare systems, such as the case of District Health Centres (DHCs) in Hong Kong.
Firstly, a comprehensive investigation into the execution of healthcare initiatives under resource constraints (e.g., limited treatments, services, staff, locations, and operating hours) in primary healthcare organisations has yet to be conducted to assess their effectiveness and performance (Pang, 2023). Secondly, the absence of a place attachment and standardised service quality in healthcare units may deter customers from utilising these facilities for health promotion and prevention (Wong et al., 2019). Lastly, individual customers have varying levels of knowledge and differing needs for healthcare treatments and services (based on factors like age, health conditions, and accessibility), and the availability of these services at healthcare units may vary (Powell & Laufer, 2010; Senic & Marinkovic, 2013; Zhang et al., 2020). These factors can significantly influence customers' overall experiences (including perceived service quality, consumption values, place attachment, and satisfaction) and their behavioral intentions towards healthcare units. Therefore, the question is how stakeholders (healthcare policymakers, healthcare providers, and customers/residents) perceive the outcomes of holistic planning, management, and operations of DHCs. It is for the future sustainability of primary healthcare systems, especially in Hong Kong.
This research is significant as it advances the underexplored theoretical development in primary healthcare service literature through a multilevel research design (Molina-Azorin et al., 2020; Senic & Marinkovic, 2013; Zhang et al., 2020). By integrating a conceptual model based on the Theory of Consumption Values (TCV) and Place Attachment Theory (PAT), this study evaluates customers' knowledge of healthcare units, their experiences with service quality attributes, consumption values, satisfaction, and behavioral intentions toward healthcare units. The multilevel research design investigates the micro (Level 1 - individual customers) and macro (Level 2 - healthcare units and Level 3 - primary healthcare initiatives proposed by the government) levels (Molina-Azorin et al., 2020). At Level 1, customers' knowledge of healthcare units could influence their satisfaction and behavioral intentions toward these healthcare units, through the quality of healthcare treatments and services offered, perceived consumption values, and place attachment (Level 2), as well as the government's healthcare initiatives (Level 3).
The Health Bureau and the Primary Healthcare Commission have proposed various initiatives (e.g., family doctors for all, Chinese medicine, and subsidy schemes) for managing DHCs as hubs for coordinating primary healthcare services for residents (DHC, 2024). Thus, DHCs could be presentable as the primary healthcare units for recruiting target samples. By employing a mixed-methods approach to achieve its research objectives, the qualitative component (Stage 1: for Level 2 and Level 3) will involve focus group interviews with 20–25 government policymakers and DHC providers, while the quantitative component (Stage 2: for Level 1) will include two surveys with 400 customers conducted through two-wave studies (Ledolter & Hardon, 2020; Ryan, 2009). Purposive and convenience sampling methods will be used to recruit respondents across DHCs for both stages. The research findings will provide valuable insights for healthcare stakeholders, aid in healthcare policy revisions, plan future DHC expansion, effectively allocate resources for healthcare services to the community, and introduce new teaching contents and curriculum design in healthcare services management.
Project Reference No.: UGC/FDS13/E04/25
Project Title: Optimized decision-making models for facilitating adaptive reuse of vacant industrial buildings for residential use
Principal Investigator: Dr THILAKARATNE Ruffina Sharmila (Chu Hai)
Abstract
The Construction Industry globally is responsible for over 30% extraction of natural resources, and 25% of solid waste generated from the linear economic model ‘take, make, dispose (Benachio et al. 2020). At present embodied Carbon accounts to 11% compared to 28% operational Carbon emissions in the construction industry. However, by 2060 embodied carbon overtakes the level of operational carbon emissions significantly due to increasing new building stock and advancements in the operational Carbon sector to mitigate impacts. Global building stock by floor area is expected to double or increase by 240 billion sqm by 2060, increasing the raw material usage and embodied carbon emissions to cater to the needs of the escalating global population (Architecture 2030, n.d.). 70% of buildings that exist today will exist by the mid-century requiring retrofitting. Adaptive reuse is gaining momentum in many countries and especially in Australia and Netherlands to address their housing shortage. By extending the lifespan of old yet structurally sound buildings, adaptive reuse promotes land use optimization, revitalization, construction waste reduction, and reduce demand on new construction material. These situations warrant even greater attention for cities like Hong Kong with limited land resources for providing adequate housing. In Hong Kong, private flatted factories represent almost 70% of industrial buildings accounting to 16.4 million sqm (HKSAR 2019). In 2023, the vacancy rate rose to 5.7% equivalent to 931,000sqm (HKSAR 2024). The revitalization policy introduced in 2009-10 led to ‘Area Assessments of Industrial Land in the Territory’ that identified 1448 existing industrial buildings within 75 estates belonging to several development zones. The largest number of industrial buildings are located in Eastern Kowloon, providing about 5.43M sqm floor space (Planning Department 2021). In Hong Kong wholesale conversion is by far the more common revitalization approach compared to redevelopment. This trend has accelerated since the 2009-10 policy introduction (Wang, 2014). To facilitate wholesale conversions, the age of the industrial buildings should be not less than 15 years, from the occupation permit issue date (Legco, 2009). Unlike new construction, adaptive reuse inherits challenges due to multiple factors influencing the decision-making process. This research aims to develop a model for establishing critical factors and optimizing the decision-making process. Artificial Intelligence (AI) / machine learning tools will be adapted for multi-factor analysis, predictive modelling, optimization and automation of routine decisions, scenario analysis and design process optimization. Although the progressive adaptation of AI in smart city developments is known, the use of AI in adaptive re-use research and implementation is still underexplored; therefore, the contribution would be significant.
Project Reference No.: UGC/FDS16/E02/25
Project Title: Multi-activation Responsive Position-agile Radiation Architecture (MAR-PARA) for Fading-free & Interference-less 6G
Principal Investigator: Prof TONG Kin-fai (HKMU)
Abstract
6G is the next giant leap for wireless technology. It promises unprecedented speeds, allowing users to download an entire 8K movie in seconds instead of minutes. Lag, those frustrating delays in online games or video calls, will become a distant memory. Furthermore, 6G promises reliable connections even in densely populated areas, ensuring everyone enjoys a seamless internet experience. This will pave the way for a world where every device, from your car and fridge to your smartphone, can effortlessly connect to the internet. However, current 5G technology, despite its impressive capabilities, faces limitations. It consumes significant energy, relies heavily on sophisticated signal processing and can become overly complex when attempting to connect a vast number of devices simultaneously on the same physical data channel. The same approach will not be sustainable 6G.
To address these challenges, this project proposes a revolutionary approach to wireless signal transmission. Instead of relying on conventional fix-positioned antennas, as seen in today's smartphones, we will develop antennas that possess the unique ability to dynamically change their shape and/or position like fluid to obtain novel spatial diversity for interference mitigation, as opposed to the conventional signal processing approach.
This innovative concept can be likened to a spotlight that can instantaneously adjust its beam to focus on different areas. By implementing this technology, we can achieve several key ad-vantages:
-
Connect more devices: similar to having multiple spotlights illuminating different directions, we can accommodate a significantly larger number of connected devices.
-
Enhance signal quality: Ensure everyone receives a strong, reliable connection, regardless of their location or the surrounding environment.
-
Reduce energy consumption: Make the entire wireless system more efficient and sustainable.
In essence, we are striving to build a smarter, more efficient wireless network that will unlock the full potential of the "Internet of Everything" as an usher in a truly interconnected world.
This project represents a significant advancement in wireless technology, with the potential to revolutionize how we communicate and interact with the digital world. By embracing this innovative approach, we can pave the way for a future where 6G technology seamlessly integrates into every aspect of our lives.
Project Reference No.: UGC/FDS13/E05/25
Project Title: Developing a Metaverse: An Immersive Digital Architecture for Enhanced Teaching and Exploration
Principal Investigator: Dr TOO Ken Wing-tak (Chu Hai)
Abstract
This research aims to enhance architectural education through the development of an immersive, mobile-friendly, browser-based metaverse platform. By integrating advanced technologies such as 3D scanning, virtual reality (VR), and augmented reality (AR), the project addresses the limitations of traditional architectural education, which often relies on static 2D images and restricted access to significant architectural landmarks.
The primary objective is to create a platform that facilitates teacher-guided virtual tours, fostering real-time interaction and engagement between educators and students. The Stanley Murray House in Hong Kong will serve as the pilot study, allowing students to explore this iconic site in a fully interactive environment. This research seeks to investigate the effectiveness of this immersive experience in enhancing student engagement compared to conventional teaching methods, particularly in architectural studies that benefit from spatial interaction.
Key components of the research include the development of digital models, user-friendly navigation interfaces, and interactive features that promote collaboration among students. The effectiveness of the metaverse platform will be assessed through quantitative surveys, utilizing Likert-scale items to measure engagement levels during virtual tours.
By exploring the potential of a metaverse for architectural education, this research aims to not only enhance learning outcomes but also contribute to the fields of heritage conservation and technological innovation. Ultimately, the findings may inspire broader adoption of immersive technologies in educational settings, providing valuable insights for educators, institutions, and policymakers.
Project Reference No.: UGC/FDS15/E04/25
Project Title: A Green Structural Efficiency Framework for New Quality Productive Forces
Principal Investigator: Dr TSANG Chun-kei (Shue Yan)
Abstract
In September 2023, President Xi Jinping of the People's Republic of China advocated for the development of “New Quality Productive Forces” (NQPF), a novel economic model centered on innovation in advanced sectors. He urged China to “lead the development of strategic emerging industries and future industries.” There are several key elements associated with NQPF:
-
Sectoral Focus: NQPF addresses the group development of specific sectors.
-
Supply-Side Emphasis: NQPF prioritizes supply-side dynamics.
-
Green Development Consideration: NQPF incorporates concerns regarding carbon emissions and other undesirable by-products.
As a relatively new concept, literature on the measurement and study of NQPF is limited. To our knowledge, no existing tools adequately address this topic. We propose that the framework of structural efficiency, within the context of productivity and efficiency analysis, is well-suited for examining NQPF. Structural efficiency measures the overall productive efficiency of a group with varying production units (PUs). However, current structural efficiency models have notable shortcomings, particularly in their neglect of undesirable outputs.
This proposed project aims to bridge this research gap by establishing a green structural efficiency framework that integrates the three aforementioned key elements to evaluate the achievement of NQPF in China's regional economies and listed firms. We will develop a comprehensive model of green structural efficiency to assess the productive efficiency of a group of PUs, taking undesirable outputs into account.
Several objectives in measuring green structural efficiency will be explored, including maximizing desirable outputs while holding other variables constant, minimizing undesirable outputs while holding other variables constant, or a combination of both. Additionally, the project will investigate the sources of inefficiency through various approaches, determining whether misallocation occurs within or among PUs. Finally, the new green structural efficiency framework will be applied to evaluate the Chinese regional economy and firms listed in China, with implications for NQPF identified.
Project Reference No.: UGC/FDS14/H13/25
Project Title: Diverse and Multicultural Shakespeare: Children’s and Young Adult Book Adaptations and Retellings in the 21st Century
Principal Investigator: Dr TSO Wing-bo (HSUHK)
Abstract
The 21st century has witnessed a transformative shift in the adaptations of Shakespeare’s works for children and young adults (YA), reflecting the emergence of diversity, equity, and inclusion ideals. This project examines how contemporary children’s and YA adaptations not only make Shakespeare’s classic texts accessible but also reimagine them through various cultural, racial, and social identities. To achieve this, the study will employ (i) a corpus-assisted critical discourse analysis and (ii) a textual analysis to examine children’s and YA adaptations and retellings published from 2000 to 2025. These methods will facilitate a comprehensive investigation into the multicultural and inclusive elements present in the texts. Part one of this project will investigate how these modern adaptations and retellings not only amplify female voices but also shed light on the nuanced intricacies and strengths of these characters. By exploring the evolution of portrayals of female characters like Ophelia in Hamlet and Lady Macbeth in Macbeth, the project aims to showcase how these adaptations and retellings gradually break free from traditional gender stereotypes, promoting a deeper understanding of female empowerment and agency. Through a closer examination of these characters and their expanded roles, the project seeks to illuminate the transformative impact of modern interpretations on reshaping perceptions of gender dynamics and fostering a more inclusive narrative landscape.
Part two of the project investigates the representation of ethnicity, particularly focusing on characters like Shylock and the Prince of Morocco in The Merchant of Venice. Traditionally depicted through negative stereotypes — Shylock as a greedy, vengeful Jew, and The Prince of Morocco as an exoticized “other” — these characters have long been controversial figures within Shakespeare’s canon. The historical portrayal of Shylock has often perpetuated harmful stereotypes about Jewish identity, reducing him to a caricature of avarice and vengeance. In contrast, the Prince of Morocco has frequently been presented as a mere exotic figure, lacking depth and individuality. Recent adaptations, however, aim to portray these characters with greater nuance and humanity. This part of the project will analyze how these modern interpretations challenge stereotypes and foster a more empathetic understanding of race and identity, encouraging young readers to think critically about the representation of marginalized groups in literature.
Part three explores how modern children’s and YA adaptations and retellings handle the concept of gender fluidity through cross-dressing the modern-day LGBTQ+ perspective. These plays feature characters such as Portia in The Merchant of Venice and Viola in Twelfth Night, who adopt male disguises to navigate a patriarchal society, raising questions about identity and societal norms. This exploration will analyze how contemporary versions interpret and present gender roles, often emphasizing the fluidity of gender and the performative aspects of identity. By reimagining these narratives, adaptations and retellings allow young readers to engage with themes of self-expression, autonomy, and the complexities of gender identity in a nuanced manner. The project will investigate how these adaptations encourage discussions around gender fluidity and bring in the queer perspective, ultimately empowering young readers to embrace a more inclusive understanding of gender.
Together, these three parts aim to provide a comprehensive understanding of how contemporary adaptations and retellings of Shakespeare can foster critical discussions about gender, race, ethnicity, and identity among young readers. By highlighting diverse representations and encouraging empathy, this project seeks to enrich children's and YA literature, making it more inclusive and reflective of modern complexities. Ultimately, the findings will empower educators and parents to use these adaptations as tools for meaningful conversations, helping to cultivate a generation that values diversity and inclusiveness in both literature and life.
Project Reference No.: UGC/FDS16/E12/25
Project Title: Frequency selective surface as energy saving smart windows.
Principal Investigator: Prof VELLAISAMY Arul Lenus Roy (HKMU)
Abstract
Hong Kong, renowned for its high-density urban landscape, faces increasing demand for innovative construction solutions to enhance energy efficiency while supporting the city’s digital infrastructure. Indoor air-conditioning dominates energy consumption during summer, driving the need for advanced smart window technologies that can block thermal radiation while ensuring optimal indoor mobile network connectivity. However, conventional energy-efficient strategies block thermal radiation effectively while impairing radio signal penetration, hindering mobile network performance in a digitally reliant economy. To address this challenge, we propose a novel approach utilising Frequency Selective Surfaces (FSS) embedded with infrared (IR) (100 – 400 THz) reflecting properties, enabling effective thermal insulation while maintaining superior mobile signal (0.5 - 30 GHz) transmission and visible (400 – 800 THz) transparency. Unlike traditional FSS technologies, which are constrained by limited bandwidth, fixed frequency response, and complex fabrication processes, our solution introduces a conformable FSS fabricated from Indium Tin Oxide (ITO)- coated glass microspheres. The proposed solution presents a cost-effective, scalable, and technologically advanced pathway to revolutionise smart window applications in Hong Kong and the Greater Bay Region. By integrating energy-saving capabilities with seamless mobile network support, this work aims to catalyse sustainable and digitally resilient construction practices for densely built urban environments. By improving energy efficiency in buildings and supporting digital infrastructure through better mobile network penetration, the project contributes to creating more sustainable and technologically adaptive cities. The proposal embodies a multi-faceted approach to sustainable development by validating the proposed energy-saving windows in collaboration with REC engineering (HK) Limited.
Project Reference No.: UGC/FDS15/H04/25
Project Title: Anticipatory grief in digital communication: Narratives in hospice care center websites and support group forums
Principal Investigator: Dr WAN Jenny Yau-ni (Shue Yan)
Abstract
This study aims to identify the linguistic features that facilitate emotional expression and connection within the hospice community by analyzing digital narratives shared on hospice care websites and support group forums. The significance of this study lies in its investigation of how language shapes the hospice care experience, particularly in the context of anticipatory grief among terminally ill patients and their carers. The Economist Intelligence Unit ranked the UK as the best country for end-of-life care and Hong Kong 20th out of 80 countries. Based on UK best practices, this proposed study aims to improve the understanding of digital platforms in end-of-life care and provide insights for developing effective hospice services in diverse cultural contexts. We will build a comprehensive text corpus of 1,000 narrative stories of at least 600,000 words from hospice centre websites and support group forums in Hong Kong and the UK. This project will result in the first corpus of its kind offering original insights into linguistic development such as similarities and differences in narrative structure, language use, and attitudinal expressions between these two cultural contexts. This study, grounded in Systemic Functional Linguistics (SFL), will examine key register variables, namely— field (topic), tenor (participant relationships), and mode (communication channel), — to understand their influence on hospice discourse. We will analyze the generic structures and grief stages of the texts for guiding information flow, and we will explore discourse semantics for how meaning is constructed through specific lexical choices. For example, we will focus on the role of conjunctions in establishing coherence, the use of personal pronouns in fostering personalization and intimacy to enhance emotional engagement, and the types of verb processes that convey mental states and actions. Key evaluative resources will also be investigated to reflect the speaker’s stance and attitude toward the experiences of end-of-life care. Using a mixed-method strategy, this project will use corpus statistical analysis methods such as Loglikelihood Ratio and AntConc to extract and analyze linguistic patterns from the corpus data. Based on our pilot study, early observations reveal that narrative structures and expressions of grief and care vary between institutional and individual narratives. Hospice care websites use more formal and structured language, while support group forums have a personal and emotive style. Hong Kong narratives emphasize familial relationships and communal support, while UK narratives may reflect individualistic perspectives. A larger data set analysis may reveal subtle differences in narrative styles and attitudinal expressions between the two regions, revealing significant differences in lexicogrammatical features, structural organization, and interpersonal meanings across digital platforms and cultural contexts. In terms of theoretical and cultural implications, the study can advance the field of SFL by applying its linguistic principles to the analysis of digital narratives in a healthcare setting, particularly the interpersonal metafunction of language in grief narratives. Comparing grief narratives from Hong Kong and the UK will contribute to cross-cultural understanding and tailored support services. Practical and policy implications include assisting health professionals, carers and support groups to understand the various ways in which people express and manage anticipatory grief in digital environments, as well as improving care support services. The research findings will be disseminated through the publication of academic journal articles, conference visits, and seminars, with the intention of stimulating further research and attention in the field of hospice care. The overall aim of this research project is to understand hospice grief and, as encouraged by the United Nations Sustainable Development Group (UNSDG), to promote the wellbeing and healthy living of an ageing population.
Project Reference No.: UGC/FDS16/E11/25
Project Title: Enhanced Data Curation and Graph-based Learning for Dual-Drug Synergy Prediction
Principal Investigator: Dr WANG Dan (HKMU)
Abstract
Background: The progression of complex diseases, such as cancer, commonly involves multiple pathways, and therefore targeting a single pathway by monotherapies often leads to low therapeutic efficacy and high toxicity. Combination therapies with synergistic effects are a well-established concept and offer a promising way to combat complex diseases. However, discovering synergistic drug combinations necessitates the exploration of a vast combinatorial space, where complete in-vitro or in-vivo examinations are infeasible. Computationally investigating this space or its subspaces, in a cost-effective way, has thus become a dynamic research domain.
Project objectives: Most computational approaches for predicting synergistic drug combinations are developed in a data-driven manner, and data is the foundation of these approaches. Large-scale drug combination screening data is a superb resource, but current databases often suffer from a lack of recent updates, limited disease coverage, and insufficient annotations. It makes enhanced data curation our first objective in this project. Secondly, graph-based learning has shown promise among key deep learning strategies in diverse fields, but its specific effectiveness in drug synergy prediction remains an active field of exploration. Seeking the possible role and capabilities of this technique in dual-drug synergy prediction is another objective of this project. Finally, a website will be created to release the curated data and the developed learning models to the public.
Main strategies: Enhanced data curation. As a primary focus of high-throughput screening techniques, the combinatorial subspace regarding dual-drug combinations tested on various cell lines will be comprehensively examined. First, cancer-related data will be gathered, filtered and standardized, with an emphasis on the most recent updates. Similar procedures will be applied to the curation of data covering well-known diseases other than cancer. Lastly, these data will be annotated and evaluated from multiple perspectives. Seeking the possible role and capabilities of graph-based learning in dual-drug synergy prediction. First, protein interaction field graphs will be generated as representations for both the drugs and cell lines, to balance the information asymmetry existing in previous representations. Then proper interaction modes will be developed to enhance the learning of those graphs, based on concepts like domain alignment and heterogeneous graphs. Additionally, promising ways to interpret the model in biology-inspired mechanisms, with respect to node feature significance and attention-driven insights, will be examined. Creating an easy-to-use website. Relying on the GitHub and Zenodo platforms, a user-friendly website will be established, to present the data in multiple angles and share the learning models with the public.
Impact of research: By meeting the above challenges, this project will output theories and practical tools for effectively exploring synergistic dual-drug combinations. It will contribute to the steady growth of precision decision domain and the health AI community, ultimately aiding the pharmaceutical industry in long-term drug discovery processes. Last but not the least, this project will benefit the undergraduate education in the local institution by increasing student engagement in research and positively influencing the current curriculum.
Project Reference No.: UGC/FDS16/E03/25
Project Title: Multimodal Context Modelling and Knowledge Fusion for Explainable Dialogue in Healthcare
Principal Investigator: Prof WANG Fu-lee (HKMU)
Abstract
Population aging has become a global issue. Due to increased life expectancy, declining birth rate, and the aging of the “baby boomer” generation, the proportion of elderly population is estimated to rise from 9.3% in 2020 to 16.0% by 2050. This demographic shift poses profound social and economic challenges, particularly in the healthcare sector. As the elderly population increases, especially among chronic disease patients, there is a growing demand for rapid, accurate, and convenient medical consultations. Chronic disease treatment often requires long-time tracing of patient medication, disease status change, self-management, information consolation, etc. This makes healthcare not only a matter of human well-being but also a core issue for societies facing the challenges of population aging.
Therefore, this research aims to explore how large language models-empowered explainable dialogue can be leveraged to improve medical treatment for chronic disease patients, thereby alleviating the healthcare pressures associated with population aging. Specifically, focusing on chronic conditions such as diabetes, we aim to design a specialized healthcare dialogue system that integrates various modalities, including text, images, audio, and other data, to build a model capable of providing reliable and accurate responses. The research also aims to reduce hallucinations (false or misleading information) in healthcare dialogues, helping patients access trustworthy health consultations. However, there are three types of challenges remains. Firstly, dialogue context representation is an essential task for healthcare dialogue since dialogue context understanding is crucial for interpreting user inquiries and facilitating subsequent dialogues. In the real world, healthcare dialogue context is typically multi-modal. Physicians need to accurately interpret these contexts to make correct diagnose decisions. Therefore, how to model multi-modal healthcare dialogue context effectively is a fundamental research problem. Secondly, healthcare dialogues require the retrieval of professional knowledge as the basis for diagnosis. Precise multi-modal knowledge retrieval is the key for guide large language models to generate explainable responses to users. Existing knowledge retrieval methods mainly focus on single modality and they cannot distinguish fine-grained differences between two similar knowledge. Therefore, how to retrieve fine-grained multi-modality knowledge precisely is also an essential problem. Thirdly, how to fuse the retrieval knowledge and dialogue context modeling for precise response generation is also a challenge. Patients generally lack professional knowledge, thus healthcare dialogue models need generate understandable multi-modal responses to users.
This study aims to leverage multi-modal large language models to address the healthcare needs of chronic disease patients in aging societies such as Hong Kong. Specifically, we will design methods for modeling multi-modal healthcare dialogue contexts to enable more accurate medical consultations. The focus will be on developing a precise knowledge retrieval model to ensure the reliability and granularity of the medical knowledge. Additionally, the study will design a fusion strategy to integrate multi-modal knowledge with dialogue context, generating explainable healthcare responses. The models developed in this research will contribute to improving the reliability of chronic disease consultations in an aging society, ensuring that users can access trustworthy medical advice more easily and efficiently. Ultimately, these innovations may also have broader applications in other domain such as education that require accurate, multi-modal information processing.
Project Reference No.: UGC/FDS15/H10/25
Project Title: Comparative Approaches to Shangshu Evidential Scholarship in Qing China and Edo Japan: The Works of Jiao Xun and Yasui Sokken
Principal Investigator: Dr WANG Li (Shue Yan)
Abstract
Evidential scholarship (Kaojuxue, 考據學), a discipline centered on textual verification and authentication, flourished in Qing-dynasty China (17th–19th centuries), particularly during the Qianlong and Jiaqing reigns, giving rise to the Qian-Jia Evidential Scholarship 乾嘉考據學 tradition. This intellectual movement later spread to Japan and Korea, dominating East Asian academia for nearly three centuries. Previous studies classify Qian-Jia scholars geographically—e.g., Wu School 吳派 (Hui Dong 惠棟), Wan School 皖派 (Dai Zhen 戴震), and Yangzhou School 揚州學派 (Jiao Xun 焦循). However, such frameworks oversimplify individual contributions and the internal complexity. In response, this project adopts a Sino-Japanese comparative perspective to explore the core characteristics of evidential scholarship.
Jiao Xun (1763–1820) and Yasui Sokken 安井息軒 (1799–1876) represent the pinnacle of mature evidential scholarship, yet their Shangshu 尚書 studies remain understudied. Critiques of the apocryphal Old Text Shangshu 古文尚書 and the forged Kong Zhuan 孔傳 marked the genesis of Qing evidential scholarship, while Shangshu studies mirrored broader trajectories of Qing intellectual history. Consequently, their Shangshu studies have been selected as the focus of this research. This project primarily analyzes research methods and underlying philosophies, aiming to make three breakthroughs:
1. Jiao Xun’s Ambiguous Affiliation. Despite openly admiring Dai Zhen and criticizing Hui Dong’s rigid Han Learning 漢學, Jiao’s Yugong Zhengzhu Shi 禹貢鄭注釋 exclusively upheld Ban Gu 班固, and his Shangshu Bushu 尚書補疏 praised the forged Kong Zhuan. This stance is typical of the Wu School. Crucially, he failed to master the Dai Zhen school’s refined method of “deriving meaning from phonology” 因聲求義.
2. Yasui Sokken’s Hybrid Approach. His Shosetsu Tekiyō 書說摘要 heavily relies on Han scholars and extensively cites Qian-Jia works, irrespective of whether they belong to the Wu or Wan schools. When conflicts arise, he favors Han views—a tendency characteristic of the Wu School. Nonetheless, his scholarship also retains stylistic elements of the Edo tradition.
3. Sino-Japanese Comparative Features. Both scholars share: (1) “Rejecting Song Learning in Favor of Han Learning” 反宋歸漢; (2) “Challenging the Old Text and Exposing the Forged Kong Zhuan” 疑古辨偽; (3) “Evidence-Based Documentary Analysis” 文獻實證. The first two points reflect shared intellectual frameworks, while the third represents evidential scholarship’s most prevalent feature.
However, neither scholar mastered the Wan School’s core method, “deriving meaning from phonology”, which Dai Zhen encapsulated as “Seek rational inference, not rigid adherence” 但宜推求，勿爲株守—termed here “Reason-Based Inference” 理據推求. The essence of this method lies in discerning semantic relationships through phonological systems. This method, alongside cultural and personal factors, distinguishes the internal differences among evidential scholarship schools.
Jiao Xun’s legacy resides in his critical reflection on textual methods and attempts to forge new avenues in Jingxue (經學, Classical Studies), though methodological limitations led to ultimate failure. Although Yasui’s intellectual grasp was less innovative, it did not hinder the further development of evidential scholarship during the Meiji period. The influx of Western knowledge precipitated modern transformations in both countries, yet evidential scholarship successfully integrated with modern disciplines, retaining enduring scholarly relevance.
As an initial phase of broader research on East Asian Shangshu studies, this project will expand to regional comparisons, revealing evidential scholarship’s adaptability across intellectual traditions.
Project Reference No.: UGC/FDS14/B06/25
Project Title: Which Relationships Are Right for Me, LMX, TMX, or Neither? The Influence of Regulatory Foci and Workflow Network on Social Exchange and Newcomers Adjustment
Principal Investigator: Dr WANG Linda Chang (HSUHK )
Abstract
The purpose of this study is to develop an integrated theoretical model of the antecedents and consequences of social exchange relationships, involving supervisor (leader-member exchange: LMX) and fellow team members (team-member exchange: TMX) in the newcomer socialization context. Wang and Hollenbeck (2019) used the colloquialism “teacher’s pet” to describe individuals who have high LMX, but low TMX. They also coined the phrase, “class clown” to characterize individuals who have low LMX but high TMX. Although these pejorative terms may not closely apply to the adult work context, from newcomer’s perspective, the need to understand who takes alternative paths of social exchange, why they do it, and how it influences newcomer adjustment outcome is clearly important for understanding team dynamics and leadership in the newcomer “on-boarding” process.
Project Reference No.: UGC/FDS24/E02/25
Project Title: Explore the Potential of Carbyne-Fused Carbon Nanotube for Upgrading Lithium-Sulfur Batteries Together with the Development of AI-Driven Aging Prediction
Principal Investigator: Dr WONG Chi-ho (PolyU SPEED)
Abstract
In the rapidly evolving field of drone technology, battery performance is paramount, directly influencing flight duration, power efficiency, and payload capacity. Drones have the potential to revolutionize various aspects of our daily lives, including transforming delivery services, improving disaster response, and conducting environmental monitoring. As the demand for efficient, reliable drone operations increases, the development of advanced battery technologies becomes essential. Among the various components, battery electrode materials play a crucial role, and the integration of high-conductivity materials holds promise for revolutionizing battery performances. Recent advancements have unveiled the potential of encapsulating carbyne within carbon nanotubes (carbyne-nanotube composites) to significantly enhance their electrical conductivity, making them a promising frontrunner for next-generation electrode materials.
As Hong Kong embarks on an ambitious initiative to foster a “low-altitude economy,” our project seeks to explore the transformative potential of lithium-sulfur (Li-S) batteries enriched with carbyne-nanotube composites. While short carbyne fragments show limited efficacy in improving the electric conductivity of the composites, the synthesis of long-chain carbyne within carbon nanotubes could lead to significant advancements in battery performance. Hence, we are pioneering innovative high-temperature fusion pathways to lengthen the internal carbyne chain. Recognizing the challenges posed by the time-intensive and empirically-driven nature of fusion experiments, we propose the development of an advanced simulation framework. This framework will allow us to predict the outcomes of the fusion process and to elucidate how internal carbyne structures influence the electrical conductivity of carbon nanotubes. By correlating these predictive models with experimental data, we aim to streamline the production of long-chain carbyne for enhanced electrode materials, subsequently testing their performance in Li-S batteries. In addition, we will leverage artificial intelligence algorithms to forecast the aging effects of Li-S batteries, ensuring the elimination of underperforming batteries in the early stages. This research not only advances carbyne manufacturing but also paves the way for integrating AI and novel carbon materials with cutting-edge battery technologies, promoting Hong Kong as a stronger competitor in technological innovation.
Project Reference No.: UGC/FDS16/M16/25
Project Title: Comparison of three Administration Routes, Bathing, Topical and Feeding of two traditional Chinese Medicines, Lonicerae japonicae flos and Scutellariae radix, on Growth, Immunomodulation and Disease Control in Nile tilapia and the underlying mechanisms
Principal Investigator: Dr WONG Emily Sze-wan (HKMU)
Abstract
Freshwater aquaculture plays a vital role in global food production, particularly in the Guangdong-Hong Kong-Macao Greater Bay Area. To maximize production efficiency, intensive cultivation is commonly implemented in the aquaculture industry. However, these intensive farming conditions create significant burden on fish populations through food competition, aggressive interactions, and aquaculture handling procedures. Such stressors lead to increased physical injuries, compromised immune function, reduced growth performance, and ultimately, greater susceptibility to various lethal pathogens, and hence, economic losses.
To address problems arising from intensive aquaculture, fish farmers currently rely heavily on chemical disinfectants such as formalin and various antibiotics to combat disease outbreaks and promote growth. However, the excessive use of these antimicrobial agents has resulted in serious environmental concerns, including the emergence of antibiotic-resistant bacteria and ecosystem deterioration. Moreover, the accumulation of chemical residues in fish tissue poses substantial risks to human health through the food chain. Therefore, there is an urgent need to develop sustainable alternatives that can enhance fish immunity and disease resistance while maintaining production efficiency.
Traditional Chinese medicines (TCMs) have shown significant therapeutic effects clinically, particularly in regulating hormonal balance, modulating immune responses, and controlling microbial infections. Research interest in TCM applications for aquaculture is increasing. Although some TCMs have demonstrated growth-stimulating, immunomodulating and infection-controlling effects, the underlying physiological and molecular mechanisms have not been fully explored.
TCM can be administered in different routes suitable for aquaculture applications, including bath immersion for external surface treatment and internal stress management, topical application for localized skin infections and wounds, and oral administration through fish feed for systemic distribution. Even when the same TCM is investigated, the most effective administration route for growth, immunomodulation and infection control has not been compared. Among potential TCM candidates, Lonicerae japonicae flos (LJF, Jinyinhua) and Scutellariae radix (SR, Huangqin) are particularly promising. These herbs, traditionally used for clearing heat and resolving toxins, have demonstrated growth-stimulating, immunomodulating, and infection-controlling effects in aquaculture. However, systematic research on Chinese medicine applications in aquaculture remains limited, particularly regarding administration routes and molecular mechanisms. Three research questions remain unanswered: (1) which administration route most effectively treats overcrowding-induced skin abrasions and infections while promoting growth and immune responses? (2) which candidate, LJF or SR, has higher efficacy for the above therapeutic responses, and are the effects of administration routes TCM-dependent? (3) what are the underlying physiological and molecular mechanisms of these therapeutic effects? In this proposal, Nile tilapia was selected as the fish model due to its economic value. Also, Nile tilapia is known for both its hardiness and aggressive territorial behavior during intensive cultivation. These characteristics make it an ideal model for studying treatment applications in aggressive edible fish species.
The aim of this study is to evaluate the efficacy of various preparation forms (herbal teabag, decoction extract, and active compound) of LJF and SR in Nile tilapia through three administration routes using bathing, topical and feeding methods on growth performance and microbial infection control. The study will also explore the underlying mechanistic pathways using the advanced proteomic and transcriptomic approaches. The findings generated from the current study will promote the translation of Chinese medicines from human system to fish aquaculture.
Project Reference No.: UGC/FDS25/E06/25
Project Title: A Study of Structural Behaviour of Prestressed Mechanically Laminated Bamboo-Concrete Composite Beam
Principal Investigator: Dr WONG Ho-fai (THEi)
Abstract
Hong Kong is a well-known international city with a substantial amount of construction work for buildings and infrastructure annually. Reinforced concrete is the most common building material; however, it is not environmentally friendly. A large amount of carbon dioxide is generated in the atmosphere during the production of Ordinary Portland Cement, which is the major component of concrete. To reduce the release of carbon dioxide, Timber-Concrete Composite Beam (TCC) systems have become popular in recent years. TCC combines concrete and timber to act as a composite structure, serving applications from simple floors to long-span bridges. However, there are growing concerns about its use, as it can induce deforestation and ecological impacts. This triggers the need for alternatives to timber for construction purposes.
Bamboo, a fast-growing giant grass, is a good candidate for replacement of wood materials due to its comparable mechanical properties and easy processing features. Mechanically laminated bamboo is one of the most common products of engineered bamboo, which attracts extensive research attention. The raw bamboo culms are first cut into sections according to the node locations. They are then split into strips, with both the inner and outer layers being peeled off. The strips are steamed for one hour and dried for two days to remove excessive moisture content. The strips are glued together and pressed under pressure on both sides, forming sliced panels. The sliced panels are then glued together under pressure to form a laminated bamboo component.
Similar to TCC, the mechanically laminated bamboo component normally works with a concrete flange, which may be cast in-situ or precast, to form a laminated bamboo-concrete composite beam (BCC). The concrete flange and laminated bamboo component are connected by shear connectors to induce composite action. It has been shown that BCC has good mechanical and ductility properties, which may replace timber as a sustainable system for construction.
Previous study by the authors has indicated that interface shear bonding strength is a crucial factor in determining the flexural capacity and structural behavior of BCC. The type of shear connector and its corresponding arrangement could be enhanced for better performance. In addition, the new concept proposed in this proposal, by using vertical prestressing at the interface is considered to be an effective measure to increase shear bonding for further improvement.
This research study aims to investigate the structural behavior of BCC under the coupling effects of shear connectors and vertical prestress at the concrete and laminated bamboo interface. Numerical simulations will be conducted using nonlinear finite element analysis. The numerical model, which will be validated by experimental results, will be used to perform parametric studies. Based on the numerical and experimental results, an analytical model will be developed for the evaluation and design applications of mechanically laminated bamboo-concrete composite beams. The study can provide practical guidelines to the construction industry for the use of laminated bamboo-concrete structures as a means of future sustainable development.
Project Reference No.: UGC/FDS15/H30/25
Project Title: Innovating Grammar Instruction: Cognitive Linguistics and Bayesian Modeling for Effective Learning
Principal Investigator: Dr WONG Ivy Man-ho (Shue Yan)
Abstract
If-conditionals pose significant learning challenges for second language (L2) learners due to their complex form-meaning relationships and traditional grammar instruction's limitations. Existing pedagogical materials often rely on simplified conditional types (Type 0–3) that do not fully align with authentic language use. This project addresses these gaps by leveraging cognitive linguistics (CL) to develop innovative instructional approaches that enhance learners’ conceptual understanding and real-world application of conditionals.
The current study will first analyse learner challenges in if-conditionals through qualitative evaluations of student writings and surveys. This phase aims to identify areas where L2 learners struggle and the extent to which current teaching methods contribute to these difficulties. Based on these findings, the study will design and implement two CL-informed instructional frameworks: Cognitive-Linguistics Inspired Pedagogy (CLIP) and Concept-Based Language Instruction (CBLI). These approaches emphasize conceptual links between grammar and meaning, providing learners with conceptual tools—such as visual diagrams—to internalize conditionals more effectively.
To assess the efficacy of these instructional models, the study will adopt a Bayesian mixed-effects modeling approach, analysing both fixed (instructional type, proficiency level) and random (individual learner variability, item difficulty) factors. Data will be gathered through acceptability judgment tests, metalinguistic knowledge tests, and syntactic priming tasks to measure improvements in both explicit and implicit knowledge. Additionally, qualitative insights from teacher and student interviews will supplement the findings, ensuring pedagogical feasibility and engagement.
By integrating computational modeling and open science practices, this project aims to advance evidence-based language education, offering scalable, replicable, and innovative teaching strategies for L2 classrooms. Findings will be shared through academic publications, open-access materials, and teacher training workshops, contributing to theoretical, methodological, and practical advancements in instructed second language acquisition.
Project Reference No.: UGC/FDS15/H07/25
Project Title: Translation of Third World literature in Communist Periodicals in the Cold War Hong Kong (1948-1969)
Principal Investigator: Dr WONG Ka-ki (Shue Yan)
Abstract
This project will investigate the translation of Third World literature in communist periodicals during the Cold War era in Hong Kong. The research period is set to be from 1948 to 1969 and will cover four periodicals under the control of the Chinese Communist Party (CCP) in the period, including two newspapers, Wen Wei Po (文匯報) and Ta Kung Pao (大公報), and two magazines, Literature Century (文藝世紀, 1957–1969) and Hai Guang Literature (海光文藝, 1966–67). It will be the first of its kind to apprehend the complete scenario of translations in communist periodicals in Hong Kong. It will utilize methodologies of periodical studies, translation studies, world literature studies, as well as knowledge of Hong Kong history and literature to explore the global outlook of communist periodicals and the characteristics of their translation. It aims to recognize their contribution to Hong Kong literature and their production of world literature, while also focusing on the agency of the Hong Kong leftists.
The project will examine how Hong Kong communist periodicals turned to Third World literature in the 1960s. It will establish the differences in translation direction between the 1950s and 1960s. “Third Worldism” arose after the Bandung Conference in 1955, where literary translations among Third World countries were initiated globally. China caught on to this trend and focused on Third World literature in the 1960s. This project will investigate if Hong Kong communist periodicals followed this trend, and if so, how it manifested. The focus will be on whether there were similarities and differences in the translation practices between the communist periodicals in Hong Kong, Chinese Mainland, and the transnational translation magazines of Afro-Asian Writers’ Bureau (AAWB) and Afro-Asian Writers Association (AAWA), all of which devoted much effort to translating the Third World in the decolonizing postwar world. It will also analyze how Hong Kong communist periodicals constructed a leftist repertoire of world literature as opposed to the right-wing, US-funded literary translation that consisted primarily of Anglo-American and European works. Case studies of Asian and Latin American literature will be identified to explore how the translation manifested Hong Kong’s connection within the region from the cultural network of communism. The “South-South” connection between Hong Kong and the Third World will demonstrate how “minor world literature” was circulated in the Sinophone region, and also contest to Hong Kong’s value as a case study in rising world literature studies.
Project Reference No.: UGC/FDS16/H26/25
Project Title: Enhancing Students' Acting Skills in the Aspect of Picturization through Codified Concepts from Chinese Traditional Opera: A Quasi-Experimental Study
Principal Investigator: Dr WONG Lai-ping (HKMU)
Abstract
Drama education in Hong Kong secondary schools predominantly incorporates methods derived from Western traditions, often beginning with training in non-verbal expressivity. However, students frequently face challenges in group performances due to limited prior exposure to drama, as nearly 80% of primary schools lack formal drama education. This absence of foundational skills affects performance quality, motivation, and engagement in drama activities. To address these issues, this research draws inspiration from traditional Chinese opera, a diverse system of theatrical performance that encompasses numerous regional styles, including Cantonese, Peking, and Kunqu opera. While these styles vary in aesthetic and performative traditions, they share an internally coherent use of codified, symbolic movements to vividly convey meaning and emotion through non-verbal expression. These movements—such as gestures representing "you," "me," "happiness," or "anger"—are central to the art form. For instance, actors might use gestures and spatial arrangements to symbolize staging in traditional Chinese opera, such as holding a whip to represent riding a horse or using two flags to depict a cart. By relying on these techniques, actors are able to create minimalistic visual narratives with little dependence on props or scenery, which is precisely one of the distinctive features of drama classes in schools.
Building on the principles of codification in traditional Chinese opera, particularly the deeply rooted traditions of Cantonese opera in Hong Kong, the principal investigator (PI) combines these codified elements with the improvisational practices of Western drama education to develop a set of techniques specifically designed for students who are new to using their bodies as expressive tools. These techniques incorporate specific hand gestures and body movements to help students convey verbal messages and emotions more effectively. Working collaboratively as a group, students use their bodies to create stage pictures for familiar scenarios they often depict, such as being in a classroom, on public transport, eating a meal, or sitting in a living room at home. Initially, students follow compositions predetermined by the teacher, but as they become more skilled, they are encouraged to introduce individual variations in quality and style. This progression enables students to make noticeable improvements in non-verbal expression, allowing them to quickly experience a sense of accomplishment.
The PI incorporated these techniques into a four-lesson learning unit, naming it the Non-Verbal Expressivity Learning Unit (NVE Unit). In the 2023-24 academic year, the PI conducted a pre-experimental study, implementing the NVE Unit in two S1 classes and assessing students' progress using the Non-Verbal Expressivity Rubric (NVE Rubric), which was also designed by the PI. The results showed significant improvement in students' ability to use their bodies expressively, increased confidence, and greater engagement in drama activities. Furthermore, the NVE Rubric was validated as a reliable assessment tool through the inter-rater agreement method (Lai Ping Wong, Tak Shun Tsin (2025). Integrating codified techniques from Chinese opera into drama education: A pre-experimental study on enhancing non-verbal expressivity in Hong Kong schools. Applied Theatre Research: Socially Engaged Performance. Intellect Discover. https://doi.org/10.1386/atr_00095_1).
Building on these initial findings, this proposed study will employ a quasi-experimental design involving five schools, with eight S1 classes in the quasi-experimental group and twelve S1 classes in the reference group. The study will evaluate the effectiveness of the NVE Unit in improving students' non-verbal expressivity through pre- and posttests scored with the NVE Rubric, alongside lesson observations, interviews, and questionnaires. Expected outcomes include enhanced non-verbal expressivity skills, increased student confidence, and further validation of the NVE Rubric as a reliable assessment tool. Additionally, this research aims to explore the adaptability of the NVE Unit across diverse educational contexts and skill levels, contributing to the development of innovative and culturally enriched approaches to drama education in Hong Kong and beyond.
Project Reference No.: UGC/FDS16/P02/25
Project Title: Development of Earth-Abundant 3d Metal (Fe, Co, and Ni) Based Catalysts with Hemilabile Phosphine Ligands and Its Applications Towards Suzuki-Miyaura Coupling Reaction and Buchwald–Hartwig Amination
Principal Investigator: Dr WONG Shun-man (HKMU)
Abstract
Cross-coupling reactions are pivotal in organic synthesis, enabling the construction of complex molecules by forming carbon–carbon (C–C) and carbon–nitrogen (C–N) bonds. Traditionally, these reactions have relied heavily on noble metals like palladium and platinum due to their high catalytic efficiency and selectivity. However, the high cost and limited availability of these noble metals pose significant economic and environmental challenges, particularly in the pharmaceutical industry where large-scale synthesis and stringent purity standards are critical.
Earth-abundant 3d metals, such as iron (Fe), cobalt (Co), and nickel (Ni), offer a sustainable and cost-effective alternative to noble metals. These metals are more readily available and less expensive, making them ideal candidates for developing green and economically viable catalytic processes. Despite their potential, the application of 3d metals in cross-coupling reactions has been limited, partly due to challenges associated with their coordination chemistry and catalytic behavior. Hemilabile phosphine ligands, which possess both strongly and weakly coordinating donor groups, present a promising solution to these challenges. This dual nature allows them to dynamically coordinate and de-coordinate with the metal center during the catalytic cycle, thereby stabilizing reactive intermediates and facilitating key catalytic steps. This property makes them particularly attractive for developing new catalytic systems.
This project aims to explore the untapped potential of combining hemilabile phosphine ligands with earth-abundant 3d metals for cross-coupling reactions. By synthesizing and characterizing a novel family of 3d metal complexes with hemilabile phosphine ligands, we seek to understand their structural and electronic properties. We will then evaluate their catalytic performance in various C–C and C–X bond-forming reactions, optimizing ligand design to enhance efficiency, stability, and reusability. Additionally, we will investigate the mechanistic pathways and ligand effects to gain deeper insights into the coordination characteristics of these systems.
The successful development of 3d metal-based catalysts with hemilabile phosphine ligands could revolutionize the field of catalysis by providing sustainable and cost-effective alternatives to noble-metal catalysts. This research has the potential to significantly impact the synthesis of pharmaceuticals, agrochemicals, and other complex organic molecules. In the pharmaceutical industry, where the synthesis of biologically active compounds often involves complex cross-coupling reactions, the adoption of these new catalytic systems could lead to more sustainable and environmentally friendly chemical processes. By reducing the reliance on expensive and scarce noble metals, this innovation promises to lower production costs, enhance the safety of pharmaceutical products by minimizing residual metal contamination, and contribute to the overall sustainability of chemical manufacturing.
In summary, this project not only addresses critical economic and environmental challenges but also holds the potential to transform the pharmaceutical industry by providing a more sustainable and efficient approach to the synthesis of essential compounds.
Project Reference No.: UGC/FDS24/E03/25
Project Title: Developing a Cloud Based Artificial Intelligent Framework for Demand Side Management of Electric Vehicles’ Charging Stations Integrating Renewable Energies and Recycled Waste Batteries
Principal Investigator: Dr WU Andrew Yang (PolyU SPEED)
Abstract
In the past several years, electric vehicles (EV) have been rapidly penetrating transportation services in a more green and intelligent way in Hong Kong. The Government of the Hong Kong Special Administrative Region (HKSAR) has actively provided policy incentives to promote EV adoption in both public and private transport sectors. In fact, the Chief Executive’s Policy Address has emphasized the promotion of EV and the charging network for two consecutive years in both 2023 and 2024. The Environment and Ecology Bureau also published “Hong Kong Roadmap on Popularisation of Electric Vehicles”, indicating that no new registrations for fuel-propelled private cars, including hybrid vehicles, will be granted after the year 2035. This commitment will benefit Hong Kong residents in various aspects, including better air quality and a better natural environment.
To support the increasing number of EVs, the development of charging infrastructure is crucial, and it is also in line with the future development direction of Hong Kong. To combat climate change and to reduce carbon emissions, renewable energies are suggested to be included in the power supply of EV charging stations. With an increasing number of new EV to be deployed for local traffic, more EV battery waste is expected in the next decade. Those batteries may not fulfil the high-standard performance for driving, but they are still in good condition for energy-storage devices. Therefore, the designed EV charging station in this proposal also suggests integrating those recycled EV batteries as energy-storage devices to synergize with local renewable energies as alternative power supply sources.
The investigators team in this research proposal aims to develop a multi-dimensional energy framework of EV charging stations integrating renewable energies and energy storage devices. Firstly, in Task I, the electric vehicle (EV)’s charging demand of pilot charging stations will be forecasted using AI and machine learning algorithms with consideration of local and regional traffic patterns and driving behaviour. Secondly, in Task II, Solar PV panels will be applied together with small-scaled wind turbines, as renewable energy sources, to generate electricity for local EV charging stations. In addition, recycled EV batteries will be collected and examined for appropriate power capacity and performance. Those selected and recycled EV batteries will be integrated into the EV charging station as alternative energy storage devices. Finally, suggestions will be provided for policy makers on the new design of EV charging stations considering renewable energies like solar energy and wind power, as well as energy storage devices using recycled EV batteries as alternative sources. Risk management issues will also be discussed.
Project Reference No.: UGC/FDS14/B14/25
Project Title: Design of A Carbon-Neutral Cold Chain E-Fulfilment Model
Principal Investigator: Dr WU Chun-ho (HSUHK)
Abstract
This research explores balancing cost, productivity, product quality, and carbon emissions in cold chain e-fulfilment, addressing the conflict between maintaining quality and achieving carbon neutrality amid growing e-commerce demands. In this research, a carbon-neutralised e-fulfilment model for the cold chain is proposed with the following components. Firstly, the temporal inventory and pick face replenishment decisions for managing cold storage facilities are modelled using deep actor-critic reinforcement learning to eliminate temperature fluctuations caused by frequent facility access, resulting in energy-efficient warehousing operations. Secondly, a novel IoT-based hybrid thermal packaging design for multi-temperature perishable products is designed and validated through full factorial experiments, where biodegradable and recyclable materials, such as expanded polypropylene (EPP), and thermoelectric cooling technologies are considered. Subsequently, the relationship between specific thermal packaging design and maximal quality assurance time for transportation can be determined. Thirdly, a combinatorial optimisation model is formulated to determine the optimal order of packing materials and delivery routes, considering green aspects, thereby ensuring the wise utilisation of packaging materials, fuel, and electricity. Lastly, the carbon footprint of refined e-fulfilment processes is calculated and standardised for benchmarking, while its carbon emissions can be analysed towards carbon neutrality.
Project Reference No.: UGC/FDS25/E09/25
Project Title: Hybrid Fiber-optic Sensing and Transient-Based Leak Detection with Artificial Intelligence for Water Distribution Networks
Principal Investigator: Dr WU Huan (THEi)
Abstract
Water distribution networks are the lifelines that deliver clean water to our homes and businesses, however, up to 30% of this precious resource is lost due to pipeline leaks over the world. Finding and fixing these hidden leaks can be extremely challenging, especially because most water pipes are buried underground and span vast areas. Our research tackles this problem by combining advanced optical fiber sensors with new signal processing and machine learning methods. Instead of installing many separate sensors, we use “distributed” fiber-optic cables that act like long, continuous sensors along the pipeline. These cables can detect tiny vibrations and pressure changes that indicate a leak – even if it is only a small crack. Moreover, we are developing novel ways to place these fiber sensors inside existing water pipes, avoiding the need for large-scale digging and road closures.
Another key part of our approach is using active pressure waves – similar to gentle “pulses” of water – sent through the pipe. When these pulses encounter a leak, they reflect in a detectable way. By analyzing these reflections, we can pinpoint the location of a leak much more accurately than traditional methods. We are also applying deep learning algorithms to make sense of the vast amounts of data collected, even when only a few actual leak events are available for training.
Ultimately, our goal is to create a more reliable, cost-effective leak detection system that utilities can implement on a broad scale. This will help reduce water loss, save money on repairs, and protect our infrastructure for the future.
Project Reference No.: UGC/FDS15/H06/25
Project Title: Why Do They Not Comply with the Law? An Empirical Study on the Foreign domestic helpers’ Illegal Hawking in Hong Kong
Principal Investigator: Dr XIAO Nancy Huina (Shue Yan)
Abstract
In Hong Kong, an international financial centre, there are challenges in regulating the problem of illegal hawking by foreign domestic helpers (FDHs) in public areas such as footbridges, parks, gardens, and streets every Sunday. Both citizens and lawmakers are calling on the government to address the issue of unauthorised selling of cooked food and other goods and services in the locations where FDHs frequently gather. FDHs’ hawking activities are illegal because they have violated hawking regulations and breached a condition of stay. To control hawking activities, the Food and Environmental Hygiene Department and other legal authorities have regularly inspected hawker stalls and regulated the operation of itinerant hawkers when they were found hawking in the streets. Whenever necessary, they took enforcement action to address any irregularities. Hawker control and law enforcement measures, however, have not deterred FDHs. Illegal hawking activities remain common at the places where the FDHs often gather. An understanding of why people comply or do not comply with the law is thus of interest to legal authorities to evaluate the effectiveness of law enforcement and social control approaches. Drawing on a theoretical framework of legal compliance (e.g., Tyler 1990; Tankebe 2009; Wenzel 2004; Kagan & Scholz 1980; Kagan et al 2011; Gunningham et al 2003; Thornton et al 2005, 2009; Nielsen & Parker 2012) and mixed research methods (including surveys, in-depth interviews, and comparative law analysis), the project aims to understand why FDHs choose to comply or disobey these laws. It will answer four research questions: (1) Do FDHs comply with the law and the enforcement in Hong Kong? In other words, what are FDHs’ coping and evading strategies in tackling legal enforcement and what are the variations of the legal compliance decisions? (2) How do various factors (i.e., perceived deterrence, legal knowledge, cost-benefit calculations, perceived legitimacy, social norms, and social identity) shape FDHs’ various decision-making on legal compliance? (3) How do institutional contexts, such as immigration policies, employment rights protection status and other supportive mechanisms (Non-Governmental Organizations, Consulates and activists), contribute to the FDHs’ legal compliance decision-making in Hong Kong? And (4) What are the implications for improving legal compliance and law enforcement in the field of hawking control in Hong Kong? This project will generate rich empirical findings to provide a comprehensive and deep understanding of the effectiveness and limitations of hawking enforcement by the legal authorities of Hong Kong. This study on understanding legal evasions by migrant workers will provide benefits to various legal authorities in Hong Kong. It will also bring a unique insight of migrant workers into the theory of legal compliance.
Project Reference No.: UGC/FDS15/H08/25
Project Title: Multi-Imperial Relations, Urban Politics, and Spatial Configuration in Treaty-Port China, 1860s-1930s
Principal Investigator: Dr YANG Taoyu (Shue Yan)
Abstract
This research project examines the multi-imperial dimensions—the intersection and juxtaposition of multiple imperialist powers—of Chinese treaty port cities during the late nineteenth and early twentieth centuries. It centers on China’s two largest treaty ports: Tianjin and Shanghai, two cities that were divided into several colonial concessions alongside the Chinese districts from the 1860s to 1940s. Historically, Shanghai was characterized by its tripartite division of governance—the British-dominated International Settlement, the French Concession, and the Chinese municipality, whereas Tianjin was home to up to nine foreign-controlled concessions (British, American, French, German, Japanese, Russian, Belgian, Austro-Hungarian, and Italian). Situated at the intersection of modern Chinese history, history of empires, and urban history, the present research analyzes how these multiple imperialisms shaped, and were shaped by, these two cities. This project fills a critical gap in the existing academic literature on modern imperialism in China by foregrounding the history of various interactions among multiple empires in the context of Chinese treaty port cities from 1860s to 1930s. While much scholarship on modern global imperial history and colonial urbanism has focused on the bilateral relationship between the colonizer and the colonized, the present study on Shanghai’s and Tianjin’s colonial pasts underscores the multiplicity, multilateralism, and multilayered trajectories at the heart of the colonial experiences of both imperialist powers and the Chinese. At the center of this research is the mutual constitution between the dynamics of multi-imperial relations and the unique spatial configuration of these two cities. On the one hand, the multiple constellations of colonial powers, as well as their interactions, produced ad hoc spatial arrangements characteristic of these port cities, delineated the contours of tangled political landscapes, and exerted significant impact on these cities’ physical landscapes. On the other hand, the side-by-side presence of colonial concessions, along with Chinese municipalities, conditioned the ways in which imperial powers operated within these urban spaces and interacted with one another. Focusing on multi-imperial interactions at several historical junctures defined by domestic and international crises, the present research project demonstrates the density and concentration of crisscrossing imperial trajectories within cities while situating Chinese colonial history within a global comparative framework.
Project Reference No.: UGC/FDS14/H05/25
Project Title: The Development of Hong Kong Cantonese Playwriting (1910s-1990s)
Principal Investigator: Dr YEUNG Choi-kit (HSUHK)
Abstract
In Hong Kong, where the primary spoken language is Cantonese whereas the written language is modern vernacular Chinese (baihua), Hongkongers are accustomed to the disparity between spoken and written forms of language. In the development of Hong Kong modern and contemporary drama, however, one notable trajectory has been the shift from writing in modern vernacular Chinese to writing in Cantonese. This project seeks to systematically document Cantonese-written and Cantonese-infused modern and contemporary plays in Hong Kong from the 1910s to the 1990s, before exploring the following questions:
First, with statistics dating from the 1910s—the era of the New Culture Movement, through the 1990s— the era of full-fledged Cantonese playwriting in Hong Kong, this paper investigates the moment when Cantonese-written plays in Hong Kong began to rise in number. That is, when did the awareness of Cantonese playwriting take root?
Second, before this wave of conscious effort, were there no Cantonese-written plays in Hong Kong? Was there an intermediary state? How can we periodize the development of Cantonese playwriting in Hong Kong’s modern and contemporary drama based on the data collected?
Third, in the different stages of Hong Kong theatrical development, what form did the use of Cantonese take in these plays? Was it clearly distinct from modern vernacular Chinese? How was Cantonese deployed in various degrees within the playwriting process?
Fourth, how does the use of Cantonese in works from different periods relate to various issues concerning the development of modern and contemporary Hong Kong drama, such as the Dialect Literature Movement in Hong Kong in the late 1940s, the phenomenon of Cantonese-infused playwriting in the 1970s and its relationship with the awareness of local identity professed in Hong Kong’s contemporary theatre, etc.?
Fifth, why did playwrights switch from writing in modern vernacular Chinese to writing in Cantonese? It seems that writing plays in Cantonese is not only about pronouncing local characteristics, but also about pursuing aesthetic goals such as the subtlety of dialogue and the musicality of the language. A question thus follows: how do Cantonese playwrights leverage the unique linguistic characteristics of Cantonese to create dramatic effects in their works?
This project, centered around the aforementioned key questions, attempts to explore the development of Cantonese playwriting in modern and contemporary Hong Kong drama from the 1910s to the 1990s. It will firstly gather and document the information from various university libraries in Hong Kong, particularly the Modern Drama Collection at the Chinese University of Hong Kong, Special Collections of Hong Kong University Library, the Hong Kong Academy for Performing Arts Library, and the archives of the Hong Kong Repertory Theatre and the Chung Ying Theatre Company, as well as the theatrical publications in the local literary magazines and cultural columns of newspapers. The goal is to identify plays that were either written in Cantonese or heavily infused with Cantonese. Based on the collected data, this project aims to clarify (i) the stages of development in Cantonese playwriting within modern and contemporary Hong Kong drama; (ii) the respective states of Cantonese playwriting in these stages, and the various ways in which Cantonese was incorporated to different extents into the writing process; (iii) the relationship between the use of Cantonese in these works and various aspects of Hong Kong’s modern and contemporary drama development.
To further investigate these questions from an “insider” perspective, the project will then involve interviews, where representatives from different sectors—such as senior actors, directors, playwrights, and drama scholars—will be invited to share their insights, as drawn from their respective experiences in performance, writing, and research, on the relationship between Cantonese playwriting and the performance, production, and development of modern and contemporary Hong Kong drama.
Project Reference No.: UGC/FDS11/E06/25
Project Title: Response header generation for empathic interactions in customer support conversations
Principal Investigator: Dr YEUNG Wing-lok (SFU)
Abstract
Short-text based social media such as Twitter (now rebranded as X) and Weibo have become popular online platforms for customers to interact with brands. Researchers analyse customer-brand conversations on these platforms and find extensive use of emotional tones (e.g., anxious, angry, grateful, etc.) by customers. In response, brands regularly flourish their replies with empathetic expressions such as “I understand your frustration.”, “We apologize for this experience.”, etc., to try to build emotional connection with customers. Finding the right expressions, however, requires understanding customers’ feelings based on what they say. This is traditionally performed by trained customer support agents with empathic conversation skills. Given an emotionally-charged customer tweet (e.g., “The Wi-Fi is down again. What a nightmare!”), an agent can offer some intangible support to the customer’s emotional state (e.g., “Really sorry about that …”) in addition to providing any tangible answer/support (e.g., “Please use this link to book a one-site inspection.”).
Machine learning researchers have been studying various deep learning approaches to modelling customer support conversations with an aim to generate empathic responses to customers. These approaches typically involve a separate step to recognise the emotions expressed in customer tweets and use them as conditions in the process of response generation. Experiments show that these approaches are effective to various degrees.
This research is concerned with yet another theoretically ground approach to modelling empathic customer support conversations on Twitter/X. Grounded in social psychology, this approach assumes that support agents play the role of caregiver and that customers have the need of feeling “attached” to the former. This need can be met by agents being sensitive to customer emotions. Agents would appear sensitive to the emotions by (a) offering help that relieve customers’ distress (b) show appropriate reactions to customer emotions. Although the latter form of “help” is intangible, it nevertheless fulfils the need of feeling attached in customers. Furthermore, we postulate that the agent’s emotional expression should match the customers in strength or explicitness during a particular exchange.
Although there have been attempts to generate empathic responses in customer support conversations, their focuses are mainly on predicting the appropriate type of emotion or intent of a response, but not necessarily the strength or explicitness of the emotion/intent involved.
We aim to incorporate emotion explicitness in a machine learning model of empathic customer support conversations which can support the generation of appropriate responses.
The proposed research extends previous work on empathic chatbots which are increasingly deployed in various services and industries including business, healthcare, education, etc. Endowing chatbots with human-level empathy has proved to be important and challenging. The aim of this research is to evaluate the efficacy of matching emotion explicitness between humans and chatbots during customer support conversations.
Project Reference No.: UGC/FDS14/P01/25
Project Title: Nonconvex Penalty Methods for Mix Sparse Optimization and Its Applications
Principal Investigator: Dr YU Kwok-wai (HSUHK)
Abstract
The role of big data has become increasingly pivotal across a wide range of fields, offering significant advantages while simultaneously posing various challenges. Mix sparse optimization has emerged as a powerful and effective approach for addressing these challenges, particularly due to its ability to capture sparse structures that reflect both inter-group and intra-group relationships. This methodology has found successful applications in various domains, including portfolio optimization, genomic association research, visual tracking, hyperspectral imaging, and differential optical absorption spectroscopy. Comprehensive empirical investigations into sparse optimization have revealed that nonconvex penalty methods often typically offer superior sparsity-promoting capabilities and enhanced robustness in sparse recovery when compared to the conventional convex ℓ1 regularization approach. Despite these advancements, the mathematical framework for mix sparse optimization, especially in relation to nonconvex penalty methods, is still developing.
In this project, our focus will be on nonconvex penalty methods for mix sparse optimization (NPMSO), which include smoothly clipped absolute deviation (SCAD), minimax concave penalty (MCP), ℓp regularization (where p < 1), and capped nonconvex penalties. We will delve into the consistency theory associated with nonconvex optimization models and the convergence theory of first-order optimization algorithms, alongside investigating the phase transition theory related to NPMSO problems and algorithms, with particular attention to applications in systems biology. Theoretically, we will introduce the mix restricted eigenvalue condition (MREC) to derive recovery bounds that encompass model error, absolute deviation, and ℓ2 recovery bounds for global minima, thereby enabling a quantitative evaluation of NPMSO stability. Furthermore, we will identify the lower bound of the sample ratio for random matrices that comply with the MREC, contributing to the development of the phase transition theory of NPMSO and clarifying its theoretical constraints in sparse recovery. On the algorithmic side, we will implement the continuation technique alongside the proximal gradient method (PGMC) and the alternating direction method of multipliers (ADMMC). By leveraging MREC, we will demonstrate their convergence towards an approximate true mix sparse solution, with the tolerance level being proportional to both the noise level and the recovery bounds, achieved at a geometric rate. Additionally, we will develop the phase transition theory for nonconvex PGMC and ADMMC within the context of random matrices. This convergence theory will provide strong theoretical backing for the sparse recovery capabilities of targeted nonconvex algorithms, addressing existing theoretical gaps in nonconvex penalty methods. Regarding applications, we plan to utilize our theoretical findings and numerical algorithms to infer the gene regulatory network of mouse embryonic stem cells, incorporating secondary structures that align with the mix sparse optimization framework. The effective application of our techniques in gene regulatory network inference will empower researchers to investigate gene regulation on a genome-wide scale in more complex model organisms.
Project Reference No.: UGC/FDS14/B17/25
Project Title: Service Robots or Human Staff? A Social Presence Perspective on the Hospitality Service Journey
Principal Investigator: Dr YU Yang (HSUHK)
Abstract
Service robots are transforming the hospitality industry by enhancing operational efficiency and reshaping customer experiences. According to Fortune Business Insights, the global market is projected to grow from $16.2 billion in 2022 to $62.35 billion by 2030. These robots are increasingly deployed for tasks such as check-in, greeting, and meal preparation—redefining service operations in hotels, restaurants, and tourist sites (Huang et al., 2021). This trend is especially relevant in regions like Hong Kong, where aging demographics and labor shortages intensify the need for automation. However, customer acceptance remains a major hurdle. Many consumers report discomfort when interacting with robots (IFR, 2023), raising concerns that robotization may erode the human touch essential to hospitality. Existing literature offers limited practical guidance, and empirical evidence on robot effectiveness is mixed: while some studies show robots outperform humans in boosting satisfaction (Yoganathan et al., 2021), others report negative impacts on purchase behavior (Luo et al., 2019). As Pitardi et al. (2022) emphasize, little is known about the unique advantages of robots over human-delivered services.
To address these gaps, we propose a framework with eight hypotheses comparing robots and human staff across three service stages including (1) pre-core (2) core and (3) post-core stages. We aim to advance theoretical understanding and provide managers with evidence-based deployment strategies. We will adopt a multi-method approach to ensure robustness and ecological validity. Field experiments will be conducted using robots in local restaurants. Secondary data from online platforms will be analyzed to examine adoption patterns across contexts. Controlled scenario-based experiments with adult participants will be used to establish internal validity. Together, these methods will generate comprehensive insights into the effective integration of service robots.
Project Reference No.: UGC/FDS15/H13/25
Project Title: Enhancing Compassion, Help-seeking Intention and Mental Wellbeing of Emerging Adults with ACEs through a Compassion-based Intervention
Principal Investigator: Dr YUEN Winnie Wing-yan (Shue Yan)
Abstract
Youth mental health is a growing concern globally and in Hong Kong. Emerging adulthood is a particularly vulnerable stage, and while psychosocial interventions exist, adverse childhood experiences (ACEs) which is a key contributor to poor mental health, remain under-addressed in local research and practice. Individuals with ACEs often struggle with shame and self-criticism, which hinder help-seeking and intensify psychological distress. Compassion-based interventions (CBIs) have shown promise in reducing shame, increasing self-kindness, and improving mental health outcomes. However, most studies are Western-based and may lack cultural sensitivity for Chinese populations, where social-oriented values shape emotional expression and help-seeking. The roles of compassion for others and openness to receiving compassion also remain underexplored in this context.
This study aims to develop and evaluate a culturally adapted CBI for Chinese emerging adults with ACEs, guided by the Resilience Portfolio Model and compassion frameworks by Neff and Gilbert. It will be conducted in three phases: Phase 1 will focus on developing and adapting intervention materials; Phase 2 will involve a pilot study with 20–25 participants; and Phase 3 will implement a randomized controlled trial of an 8-week intervention with 126 participants, comparing the CBI group to a waitlist control. The programme will include six 2-hour group sessions and two practice sessions to enhance participants’ emotional awareness, compassion, and internal resources to address difficult emotions. Pre-intervention, post-intervention, and three-month follow-up assessments will be conducted using validated scales and interviews to examine the changes in participants’ levels of compassion, shame, resilience, mental help seeking attitude and intention, as well as mental health indicators.
By generating evidence on the impact of CBI in addressing the psychological effects of ACEs, this study will contribute to the development of culturally responsive mental health practices. Findings will inform practice and contribute to a culturally adaptive framework for enhancing resilience and emotional well-being in young adults.
Project Reference No.: UGC/FDS14/E02/25
Project Title: Towards Robust and Cost-Efficient Vertical Federated Learning
Principal Investigator: Dr ZHANG Chen (HSUHK)
Abstract
Vertical federated learning (VFL) is a privacy-preserving machine learning framework that enables multiple parties with different feature spaces to collaboratively train a model without sharing their raw data. For instance, an insurance company could collaborate with several banks to develop a fraud detection model. This collaboration enables the model to benefit from the comprehensive knowledge of all parties involved while keeping each party's data localized.
To achieve efficient VFL, the initial task is private entity alignment, which requires pinpointing the common sample ID intersection among all parties without revealing any additional information. A practical private entity alignment protocol should be both robust and cost-efficient. However, existing protocols struggle with significant overhead challenges, especially when the number of participants is large. Moreover, current designs do not adequately accommodate participant dropouts, an issue frequently encountered in VFL due to geographical diversity and network heterogeneity among participants. Thus, it is crucial to bridge the gap by designing a cost-efficient multiple-party private entity alignment protocol for VFL systems that is tolerant of participant dropout during the alignment process.
In addition, the distributed learning architecture of VFL presents significant challenges to its robustness. If attackers control some participants and cause them to upload poisoned model updates during training, the performance of the collaboratively trained model can significantly deteriorate. Unlike horizontal federated learning, where participants transmit model updates, VFL involves sharing local model outputs. These outputs are abstract and reveal little about the specifics of the local models, making it difficult to detect malicious activity. Existing defenses are designed to increase tolerance to abnormal data and protect against data poisoning attacks. However, these designs are inadequate against sophisticated attackers who can directly manipulate the local model training process, known as model poisoning attacks, instead of merely altering the training datasets. There is an urgent need to design more robust defense schemes that can effectively detect malicious participants and defend against untargeted model poisoning attacks.
The principal goal of this project is to fill the above-mentioned research gaps and design a robust and cost-efficient VFL framework. There are three main tasks in this project: 1) Design a robust and cost-efficient multi-party private entity alignment protocol that enables multiple parties to efficiently compute data intersections without revealing dividual samples, while also accommodating participant dropouts during the alignment process. 2) Design defense schemes to protect the proposed VFL framework against untargeted model poisoning attacks. The proposed design should be able to efficiently detect malicious parties during the training process. 3) Develop a prototype system of the proposed robust and cost-efficient VFL framework to facilitate performance evaluations. Evaluate the proposed privacy-preserving entity alignment protocol across diverse datasets and party scales, and assess the robustness of the proposed defense mechanisms against various poisoning attacks in different network environments.
Project Reference No.: UGC/FDS16/H40/25
Project Title: GPT-powered problem-solving assistant for parents of children with Autism Spectrum Disorder: A pragmatic randomized controlled trial
Principal Investigator: Dr ZHANG Wen (HKMU)
Abstract
The World Health Organization reported that Autism Spectrum Disorder (ASD), which is among the most prevalent and significant developmental disorders affecting children, is diagnosed in roughly one out of every 100 children around the world. Parenting a child with ASD is a continuous and demanding journey, often accompanied by higher levels of depression and parenting stress compared to parents of typically developing children. Despite growing attention to the mental health needs of parents of children with ASD, existing interventions continue to lack accessibility and personalization.
Problem-solving skills are a stress-buffer for parental mental health. Enhancing problem-solving skills can empower parents to adapt to their daily challenges associated with their child’s diagnosis and experience reduced distress. Problem-solving skills training is an evidence-based intervention to improve problem-solving skills and reduce depression and parenting stress. Advanced technology, such as chatbots, can significantly enhance access to PSST for parents of children with ASD. The introduction of large language models (LLMs), such as OpenAI Generative Pre-trained Transformer (GPT) models, marks a transformative advancement in artificial intelligence (AI)-driven chatbots, offering capabilities far beyond those of earlier AI tools. Hence, we propose to develop a GPT-powered problem-solving assistant to promote mental health in parents of children with ASD.
Through this proposed project, we will first determine whether the tool is effective in reducing depression and parenting stress, improving problem-solving skills and well-being, and enhancing parental and children’s health-related quality of life. The results will provide evidence for the use of the chatbot in parents, inform evidence-based practice to provide mental health support for parents of children with ASD, and provide essential guidance for practitioners (such as nurses and social workers) and policy makers to improve the quality of life for families of children with ASD.
Project Reference No.: UGC/FDS24/B10/25
Project Title: A Real-time Collaborative Multi-agent System for High-frequency Nowcasting of Tourism Demand in Hong Kong
Principal Investigator: Dr ZHANG Xinyan (PolyU SPEED)
Abstract
The growing uncertainty of demand makes high-frequency nowcasts more valuable for fast decisions in policy or business decision-making in sustainable destination management. High-frequency nowcasting, however, is challenging. It is a complex and dynamic issue with high levels of uncertainties. There is a lack of studies in existing literature that generate higher-frequency nowcasts and not many real-time indicators have been incorporated in nowcasting research. Taking stock of collaboration theory, complexity theory, and actor-network theory, this proposed study is among the first attempts to address the high-frequency nowcasting challenge by exploiting the merits of advanced forecasting models, artificial intelligence, multi-agent system (MAS), cloud-based computing, and the real-time Delphi (RTD) method to address the issue. Multisource real-time indicators will be included to generate high-frequency tourism demand nowcasts.
Collaboration theory and complexity theory suggest that collaboration is a critical driver for achieving higher nowcasting accuracy in complex and dynamic systems through more complex models. Based on these theories, this project proposes a collaborative multi-agent architecture and will build a cloud-based nowcasting system prototype to validate it. Following the principle of actor-network theory, the proposed multi-agent system (MAS) consists of autonomous individuals (or agents) who work toward goals (e.g. solving emergent issues) in dynamic and complex sociotechnological networks. MAS has proven to be effective as an approach when heterogeneous agents are collaborative for integrating information or knowledge to achieve better performance.
It is expected that this study will not only generate theoretical and methodological contributions to tourism demand forecasting research, decision theory and behavioral operations management research, but also provide practical implications to various tourism stakeholders including the public sector such as Hong Kong Tourism Board, trade associations such as Hong Kong Hotel Association, tourism industry practitioners (e.g. transportation, hotels, attractions, etc.) in policy formulation and decision making. This study will also offer substantial implications to software vendors for the design of behaviorally-informed FSSs to achieve human-algorithm integration and overcome algorithm aversion in real-world forecasting environment.
Project Reference No.: UGC/FDS51(25)/E02/25
Project Title: Optimizing Interventions for Mild Cognitive Impairment: Exploring the Impact of Robotic Therapy and Delivery Modality on Cognitive Outcomes
Principal Investigator: Dr ZHONG Junpei (THEi)
Abstract
This research project aims to investigate the best ways to improve cognitive function in individuals experiencing Mild Cognitive Impairment (MCI), a condition affecting a growing segment of the aging population. MCI involves slight but noticeable changes in thinking abilities and can sometimes precede more serious conditions like Alzheimer's disease. Early intervention can significantly impact long-term cognitive health.
Our focus is on personalized brain training, leveraging cutting-edge technology to optimize cognitive interventions, i.e. using a robot to replace humans to do the therapy. We hypothesize that different types of therapy might be more effective for different individuals, and the delivery of the therapy plays a crucial role. We will compare two promising approaches: (1) traditional cognitive training, involving exercises to strengthen mental skills, and (2) robot-assisted therapy, utilizing interactive robots for personalized feedback and motivation.
We will use functional near-infrared spectroscopy (fNIRS), a safe brain imaging technique, to understand how these therapies affect brain activity. This will provide valuable insights into the neural mechanisms underlying cognitive improvement.
Working with individuals diagnosed with MCI, we will tailor training programs to their specific needs. By comparing the effectiveness of both approaches, we aim to identify the most effective and engaging strategies for improving cognitive outcomes.
This research has the potential to transform how we approach early memory problems, leading to improved quality of life for individuals with MCI, their families, and caregivers. The innovative use of robot-assisted therapy and fNIRS brain imaging will contribute significantly to the field.
Project Reference No.: UGC/FDS13/E06/25
Project Title: Towards Graph-Based Learning System Under Diverse Graph Homophily: Method, Vulnerability and Robustness
Principal Investigator: Dr ZHU Yulin (Chu Hai)
Abstract
Graph-based learning methods are powerful AI tools that excel at learning from relational data, like social networks, recommendation systems, transportation networks, and transaction networks. However, they face significant trust issues because they are vulnerable to malicious adversarial noises and often get confused by task-irrelevant patterns, limiting their applications in the complex real-world scenarios. While previous research has improved trustworthiness by enhancing generalization and robustness, these efforts primarily relied on the "homophily" assumption—where similar nodes are tended to be connected—which works well for some graphs but fails dramatically for others where connected nodes are distinct (heterophilic graphs), narrowing their practical application. To address this problem, our preliminary studies have uncovered a key insight: the similarity between aggregated node embeddings can prominently distinguish reliable connections from task-irrelevant ones, providing a new foundation for graph data augmentation that works beyond homophily. Building on this, we aim to develop a more universal and automated trustworthy graph learning system that can handle various challenges—including noisy data, different learning settings, and diverse graph types—to ultimately deliver robust and reliable predictions across real-world scenarios.