Microalgae can convert light energy into measurable electrical current, making them promising living biomaterials for bioelectronic applications such as sustainable power generation, environmental sensing, wearable devices, and biosensors. However, different applications require different electrical outputs, environmental tolerances, and growth conditions, while relevant information is currently scattered across individual studies.

This project aims to make microalgae easier to identify and evaluate for specific bioelectronic applications. It combines an experimentally characterized library of freshwater microalgae collected at DKU with a literature-based database containing information on taxonomy, habitat, growth conditions, photosynthetic performance, environmental tolerance, and electrical output.
The collected data will support a machine-learning-assisted platform that allows researchers to specify an intended application, desired electrical-output range, and operating conditions and receive ranked candidate microalgae. Predictions will be experimentally validated using DKU strains.

By integrating biological characterization, electrical measurements, machine learning, and access to living microalgal strains, this project seeks to bridge the gap between microalgal research and practical bioelectronic device development.

ChatDKU is an institutionally supported, open-source large language model platform already deployed at Duke Kunshan University. This project will use ChatDKU as a real-world research testbed for an undergraduate team to build and evaluate a new LLM agent addressing one clearly defined campus need, such as academic support, career development, research discovery, or student services.

The central research question is whether an agentic system combining retrieval, tool use, planning, source verification, and calibrated abstention can complete complex campus tasks more accurately and reliably than a conventional retrieval-augmented generation (RAG) chatbot while satisfying privacy, security, and access-control requirements.

The student team will lead the literature review, problem formulation, system design, implementation, experimentation, and dissemination. Students will first identify a focused campus use case and formulate a testable research hypothesis. They will then build a conventional RAG baseline, implement the proposed agent on the ChatDKU framework, and develop a reproducible bilingual evaluation benchmark. The system will be evaluated in terms of factual accuracy, citation correctness, information freshness, task completion, robustness, latency, usability, and privacy and security risks.

Depending on the selected use case, the project may investigate multilingual interaction, automatic answer verification, prompt-injection defenses, multi-agent collaboration, or privacy-aware agent design. The project gives students an opportunity to contribute to a deployed open-source AI platform and complete an end-to-end research cycle from problem formulation and system development to rigorous evaluation and academic dissemination. Strong contributors may also have opportunities to continue collaborating with the ChatDKU research group on follow-on research projects.

Any campus pilot will proceed only after the relevant technical, privacy, data-governance, security, stakeholder, and human-subject approvals.

This project addresses the systemic neglect of janitorial staff, where frontline workers-mostly migrant employees in their 50s and 60s-face inadequate rest spaces, wage suppression, health hazards, and cultural invisibility. Despite the institution’s global reputation, janitorial staff are relegated to poor working conditions. Based on preliminary interviews and observations, this staff is cramped in utility closets for meals and breaks, prohibited from sitting or resting in public areas under threat of being labeled “slackers.”

The proposed initiative, “Restoring Dignity,” adopts a two-stage approach. First, it maps existing rest stations and advocates converting underused storage rooms into proper break areas with seating, water, and climate control. Funds will be used for a preliminary resting area, and perceptions and effects will be evaluated before and after.

Second, it launches a campus awareness campaign-including photography exhibitions, oral history interviews, and student pledges-to render this labor visible and cultivate respect, while integrating the issue into service-learning curricula.

Implementation proceeds through research, pilot interventions, and scaling, with the goal of transforming both physical conditions and institutional culture. Grounded in the principle that basic labor rights are not charity, the project aligns with broader policy trends while challenging the structural and cultural forces that render essential workers invisible. By simultaneously addressing spatial injustice and social recognition, it aims to ensure that those who maintain the campus can do so with dignity, rest, and respect.

Spectroscopy is a powerful technique for studying and identifying materials, but commercial spectrometers are often expensive and complex. This project aims to design and build a simple, low-cost, and compact spectrometer using accessible optical and electronic components. Students will explore the basic principles of spectroscopy and develop a working prototype through hands-on design, fabrication, assembly, calibration, and testing. The project will involve optical alignment, detector integration, basic electronics, data collection, and spectral analysis. The prototype will be tested using several well-defined reference samples, such as polystyrene, silicon, and selected simple organic materials, to evaluate its ability to detect and distinguish characteristic spectral features. The project has both academic and practical value. Students will apply fundamental concepts in optics and spectroscopy to a real scientific instrument while gaining experience in experimental design, fabrication, testing, and data analysis. The project will also investigate how a low-cost optical system can achieve useful spectral measurements compared with established laboratory instruments. The project is intentionally limited in scope and designed to be feasible within one year. The main stages will be system design, prototype fabrication, calibration, measurement, and performance evaluation. The expected outcome is a functional proof-of-concept spectrometer and a documented evaluation of its capabilities and limitations.

As embodied AI moves from digital models into robots, factories, logistics systems, healthcare, and public services, its environmental responsibility becomes distributed across global networks of critical materials, semiconductors, computing infrastructure, manufacturing, transportation, operation, maintenance, and end-of-life recovery. Existing assessments often estimate a total environmental footprint but do not clearly explain who is responsible for which impact, what evidence supports that judgment, or how different rules may change organizational behavior.

This project asks:

How can environmental responsibility in embodied-AI global supply chains be recorded, verified, and shared in ways that improve both accountability and incentives?

Students will select one clearly bounded embodied-AI system, map its principal actors, resources, decisions, and environmental effects, and build a provenance-aware dataset from credible public evidence. They will compare at least three responsibility-allocation mechanisms-for example, direct operational control, lifecycle contribution, economic value, or capacity to prevent harm-and examine how these rules may affect disclosure, procurement, production, mitigation, and responsibility shifting.

The central Minimum Viable Product will be a functional Environmental-Responsibility Ledger and Mechanism Lab. Its interactive interface will make supply-chain relationships, evidence sources, assumptions, uncertainty, missing data, and alternative responsibility allocations visible and open to comparison.

The project will incubate students as scientists, entrepreneurs, and philanthropists: conducting interdisciplinary research for international dissemination; translating the findings into a responsible startup demonstration; and returning knowledge to society through an open website, stakeholder workshop, tutorial, and multimedia educational resources.

The project will make substantive contributions to:

• SDG 9 – Industry, Innovation and Infrastructure, by advancing responsible and sustainable embodied-AI infrastructure;
• SDG 12 – Responsible Consumption and Production, by improving lifecycle transparency, responsible procurement, and circular supply-chain thinking;
• SDG 13 – Climate Action, by connecting verifiable environmental evidence with incentives for mitigation;
• SDG 17 – Partnerships for the Goals, by bringing together

students, researchers, industry, policymakers, sustainability experts, and international-standards contributors.

Local heritage is rooted in specific places, communities, and memories, yet it can also become part of a global conversation. This project explores how multilingual communication, data visualization, interactive media, and agentic technologies can help local communities share their heritage with global audiences while preserving cultural context, community authority, and sustainable local value.

Students will work with one clearly bounded museum, cultural institution, heritage community, neighborhood, or public-interest organization in Kunshan. Using maps, timelines, images, oral histories, community narratives, cultural records, and spatial-temporal data, they will create a multilingual and provenance-aware cultural collection.

The main Minimum Viable Product will be a Culturally Aware Agentic Living Atlas - an interactive and evolving knowledge system that connects people, places, cultural practices, environments, and community memories over time. It may include multilingual maps and stories, cultural archives, transparent sourcing and permissions, community review and correction mechanisms, and agentic tools for translation, cultural exploration, source checking, and personalized learning.

The atlas will also create pathways for global citizens and digital nomads to move from Visitor, Learner, Contributor to Trusted Collaborator, contributing skills, knowledge, networks, or creative work in response to community-defined needs. Students will compare different approaches to matching, reviewing, recognizing, and sharing the benefits of these contributions while examining their effects on cultural integrity, inclusion, community workload, local value, and long-term engagement.

Agentic technologies will support exploration and collaboration without replacing human judgment or community authority. Local partners will retain control over cultural meaning, permissions, sensitive materials, corrections, and responsible reuse.

The project integrates three interdisciplinary areas: Open Science, Data Governance, and Responsible Agentic Systems; Scientific Communication for Emerging Technologies; and Human-Centered Spatiotemporal Intelligence. The initial pilot will focus on one feasible Kunshan partnership, while its design may later be adapted to museums, heritage routes, cultural communities, and World Heritage sites.

The project contributes primarily to SDG 8, SDG 11, SDG 16, and SDG 17 by promoting responsible cultural entrepreneurship, sustainable communities, transparent cultural-data governance, and partnerships between local communities and global participants.

Real-time voice communication is now the default medium for work and study, yet its processing stack focuses mainly on the communication channel, addressing problems such as noise, echo, and packet loss. The speaker's own voice is largely left untouched. When platforms do intervene, the available controls are usually coarse, relying on fixed filters or signal-processing transformations derived from the source-filter model of speech production. These methods modify acoustic attributes frame by frame without explicitly representing what is being said or who is saying it. The more difficult problems, however, often lie not in the communication channel but in the speaker. A student may join an 8 a.m. seminar only minutes after waking up. Her voice is low, thick, and imprecise, and before she has finished her first sentence, everyone in the room can tell that she has just gotten out of bed. She may then stop volunteering and wait until she sounds more awake, by which time the discussion has already moved on. An instructor who has spent three hours teaching back-to-back classes may become hoarse and start swallowing the ends of sentences. Students joining remotely, already dealing with compressed audio, may begin to miss entire clauses. Similar problems arise in professional communication. A negotiation may run long, and someone's patience may become audible through a clipped, sharp, or rising tone.

Although the words themselves may be reasonable, it is often the tone that others remember, and the following conversation may be spent repairing the interaction. A researcher presenting to overseas collaborators may have to choose between speaking a second language haltingly, which can undermine confidence and authority, or relying on a synthetic interpreter voice that sounds like no one. Someone returning a call from an unknown number may prefer not to provide a stranger with a clean recording of their natural voice, yet the available alternatives often make them sound obviously artificial, which can itself create suspicion. These situations illustrate a broader limitation of today's voice communication systems. Existing tools primarily improve the transmission channel and, in limited cases, protect speaker identity by replacing the voice altogether. They provide little ability to selectively control different aspects of how a person sounds while preserving the aspects that should remain unchanged.

This project aims to address this limitation by building a real-time voice communication system that disentangles speech into four largely independent factors: linguistic content, speaker identity, prosody and emotion, and voice-quality state. The system will allow any one of these factors to be modified on demand while preserving the others. Integrated directly into phone calls and online meetings, such a system could help a user sound clearer when tired or hoarse, regulate how emotion is conveyed without changing the intended message, communicate across languages while preserving their vocal identity, or protect speaker identity without sacrificing naturalness. Rather than treating the voice as a fixed signal that must either be transmitted unchanged or replaced entirely, the proposed system treats speech as an editable representation. It gives speakers real-time control over how they are heard while allowing them to remain recognizably themselves.

Research Focus Office-based knowledge workers frequently experience brief periods of frustration, tension, fatigue, or anxiety during the workday. However, when these emotions arise, workers often have only a few seconds or minutes available and may not be able to leave their desks, open a well-being application, or complete a structured relaxation exercise. This project explores an alternative approach: a small, discreet interactive object that remains on the user’s desk and supports brief, low-effort emotional regulation during everyday work.

The project addresses two research questions:

• RQ1: How should a lightweight desktop object be designed to help office-based knowledge workers regulate momentary negative emotions during the workday?

• RQ2: How do knowledge workers use such an object within ordinary work, and what forms of interaction and feedback best support brief, low-effort regulation? The project will investigate how workers currently respond to momentary emotional strain and explore different forms of tangible interaction, such as touching, pressing, holding, or manipulating an object, combined with subtle feedback such as light or physical movement. The goal is not to develop a clinical mental-health intervention. Instead, the project focuses on designing an everyday object that allows workers to briefly pause, redirect attention, and regulate momentary emotions without significantly interrupting their work.

Academic Value The project contributes to research in human-computer interaction, tangible interaction, and digital well-being. Existing workplace well-being technologies often rely on applications, structured exercises, or interventions that require users to deliberately stop working. Less is known about how physical interactive objects can support very brief moments of self-regulation while remaining integrated into everyday work routines. By studying both users’ immediate experiences with tangible interactions and their longer-term use of the object at their actual desks, the project will provide knowledge about how physical interaction can support well-being in real-world workplace contexts.

Innovation The project has three main innovations. First, it explores an interaction form beyond screens. Rather than adding another mobile or desktop application to an already screen-intensive work environment, the project investigates physical and tangible interaction. Second, it emphasizes lightweight and discreet interaction. The object will be designed for interactions lasting seconds or minutes, requiring little attention and remaining socially appropriate in shared workplaces. Third, it follows an evidence-driven product innovation process. The final object will not be predetermined. Its physical form, interaction mechanism, and feedback will emerge from user research, concept evaluation, hands-on testing, and real-world deployment.

Feasible Research Plan The project will follow a four-stage user-centered design process:

Stage 1 – Explore User Needs: Conduct workshops with office workers to understand common emotional situations, existing coping behaviors, and unmet needs for brief workplace recovery.

Stage 2 – Evaluate Design Concepts: Conduct an online questionnaire with a larger sample to compare alternative concepts and investigate preferences regarding size, appearance, interaction, feedback, portability, workplace visibility, and privacy.

Stage 3 – Prototype and Test: Develop or select several physical concepts for hands-on evaluation. Participants will compare different forms of physical interaction and subtle feedback in a desk-like environment.

Stage 4 – Develop and Deploy: Develop an original interactive desktop prototype based on the findings and deploy it on participants’ actual desks for one to two weeks. Feedback from field use will inform iterative improvements to the prototype.

The project is feasible for Dachuang while offering meaningful technical and research challenges. Students will participate in user research, product and interaction design, physical prototyping, embedded technology development, field evaluation, and data analysis, experiencing the complete process of transforming a user need into an evidence-based interactive product.

1.Research Focus

Many people recognize the value of journaling but struggle to maintain the practice. Common barriers include not knowing what to write, finding it difficult to start, feeling pressure to write something meaningful, and losing motivation after missing several days. Generative AI creates new possibilities for addressing these barriers. Instead of simply providing a blank page, an AI journaling system can initiate a conversation, ask appropriate questions, respond to users’ experiences, and help organize their reflections. This project asks a focused research question: How should interaction with AI be designed to help people sustain everyday journaling?

The project will particularly investigate two interaction mechanisms:

• Interaction initiative: How does user-initiated versus AI-initiated interaction affect people’s willingness to start and continue journaling?

• Interaction modality: How do text-based and conversational/voice-based interactions differ in the effort required to express and record everyday experiences?

An AI journaling application with different interaction variants will be developed to investigate these questions. The variants will share the same underlying platform and AI capabilities but differ in how journaling interactions are initiated and conducted. The study will examine three key aspects of sustained journaling: initiation friction, expressive effort, and recovery after interruption. Rather than measuring only whether users like the AI, the project will study actual behavior, including how often users complete entries, how quickly they start, whether they abandon sessions, and whether they return after missing a day.

2.Academic Value

The project contributes to research in human-computer interaction (HCI) and human-centered AI by examining how interaction design affects sustained engagement with AI. Existing AI journaling systems often emphasize what AI can generate or summarize. This project instead focuses on how people and AI should interact over time. The findings can provide new knowledge about initiative, modality, user agency, and sustained engagement in long-term human-AI interaction. The research is also relevant beyond journaling. Understanding when AI should proactively initiate interaction and which modality requires less user effort can inform the design of AI companions, personal assistants, reflection tools, and other proactive AI systems.

3.Innovation

The project has three main innovations. First, it shifts the focus from AI-generated content to AI interaction mechanisms. Rather than asking whether AI can write or summarize diary entries, the project investigates how AI should interact with users to support an ongoing reflective activity. Second, it systematically investigates interaction initiative and modality. Different variants of the same AI journaling system will enable comparison of user-initiated and AI-initiated interaction, as well as text and conversational/voice interaction. Third, it adopts a behavior-centered longitudinal evaluation. Instead of relying mainly on one-time satisfaction ratings, the study will examine actual journaling behavior over several weeks, including initiation, completion, interruption, and return.

4. Feasible Research Plan

The project will follow four stages:

Stage 1 – Design and Development: Review related research and develop a mobile AI journaling application with different interaction variants.

Stage 2 – Pilot Testing: Conduct usability testing to improve the interaction flow, AI responses, voice/text experience, and data logging.

Stage 3 – Longitudinal Field Study: Deploy the system with adult Chinese-speaking users for approximately three weeks. Behavioral interaction logs, questionnaires, and interviews will be collected to understand both what users do and why different interaction mechanisms help or hinder journaling.

Stage 4 – Analysis and Dissemination: Analyze behavioral and qualitative data, derive design recommendations, and prepare a research report and publication. The scope is appropriate for a student-led Dachuang project and combines generative AI, mobile development, interaction design, user research, and data analysis. Students will experience the complete research cycle from identifying a research problem and developing an AI system to conducting a field study and communicating research findings.

Governments increasingly use AI-assisted chatbots to answer questions, process service requests, and communicate with citizens. Yet we know much less about whether interacting with an AI government representative changes what citizens are willing to communicate and how they perceive public institutions. This interdisciplinary, student-led project will develop a simulated municipal-service chatbot and conduct a randomized experiment comparing AI and human service agents using procedural or empathetic communication styles. Participants will respond to realistic urban-service scenarios, allowing the research team to examine both communication behavior and institutional perceptions.

Outcomes will include message detail and tone, willingness to express dissatisfaction, willingness to continue the interaction, feeling heard, perceived responsiveness and procedural fairness, trust in government, political efficacy, and willingness to use official participation channels. Combining public policy, political behavior, human–computer interaction, and human-centered AI design, the project will train students in literature review, research ethics, experimental design, chatbot interface development, natural-language processing, statistical analysis, and policy communication.

Expected outputs include a functional research prototype, a preregistered experimental design, an anonymized research dataset and analysis code where ethically permissible, a research report or submission-ready manuscript, a bilingual policy and design brief, and a public project presentation. The project’s innovation lies in treating a government chatbot not simply as a service-delivery technology but as a new institutional interface between citizen and state. Existing evaluations frequently emphasize accuracy, adoption, efficiency, or satisfaction. This study instead investigates whether the identity and communication style of an AI representative change citizens’ own communicative behavior: how much they disclose, how directly they criticize government performance, whether they feel heard, and whether they remain willing to use official communication channels. The project also innovates methodologically by combining a controlled randomized experiment, behavioral analysis of citizen-generated messages, post-interaction survey measures, usability testing, and basic natural-language processing. Its interdisciplinary supervisory structure ensures that neither the political consequences nor the technical and experiential dimensions of the chatbot are treated as secondary.

Environmental problems are increasingly communicated visually, through photographs, maps, infographics, data visualizations, short videos, personal stories, and social media. These formats can make complex issues easier to approach, but what attracts attention is not necessarily what people understand or trust. This project asks a simple question: How does the way we present environmental information shape what young people understand, trust, and find relevant?

Students will explore this question through one environmental issue connected to DKU and the wider Jiangsu/Yangtze River Delta region, such as extreme heat, water quality, plastic and waste pollution, or urban biodiversity. The team will choose the final case after reviewing its local relevance, the available scientific evidence, and how the issue is already being communicated. This will give students the opportunity to investigate a broader question about environmental communication through a problem that is tangible in the region where they live and study.

The project will begin by looking at existing communication. Students will conduct a focused literature review and collect a manageable sample of materials related to their chosen issue, such as news coverage, public-information materials, photographs, maps, infographics, and social-media content. They will then analyze how these materials communicate environmental problems: What kinds of evidence do they use? Do they rely on data, personal stories, images of particular places, or emotional appeals? How do they communicate risk, uncertainty, responsibility, or possible solutions? Building on this analysis, students will select a few contrasting ways of presenting the same environmental information and examine how young adults respond to them.

They might, for example, compare a data-centered approach using maps or graphs with a human-centered approach built around personal experience or a place-centered approach showing environmental change in a recognizable setting. Subject to DKU IRB approval, a small pilot study within the DKU community will explore differences in understanding, credibility, relevance, and interest in learning more.

The aim is not to identify one universally “best” format, but to understand what different forms of communication do well and where they fall short. The final stage will bring the research back into practice. Students will use their findings to create a small set of environmental communication prototypes.

If one approach makes scientific evidence particularly clear while another makes an issue feel more immediate, for example, they can experiment with combining those strengths. The prototypes can then be tested on a small scale and refined. In this way, the project moves from studying communication, to testing it, to creating something new based on the evidence. Rather than asking students simply to produce an environmental campaign or design attractive materials, the project asks them to investigate what makes environmental communication meaningful and credible before deciding what to create.

Students will be involved throughout the research process: choosing the environmental case, refining the research questions, reviewing the literature, collecting and analyzing materials, developing the audience study, interpreting the findings, and creating the final outputs. I will supervise the research design and provide guidance in environmental and science communication, media analysis, qualitative and mixed research methods, research ethics, and the interpretation and presentation of findings.

The project is designed to be ambitious but manageable within one year. It focuses on one environmental issue, a relatively small body of communication materials, a pilot audience study, and a limited number of final prototypes. Most of the work can be carried out at DKU using publicly available materials and standard research tools. By the end of the project, students will have completed a full research process while producing both original findings and practical communication materials that could be shared within the DKU community and potentially developed further in future research.

The proposed Muddy Water Coffee research project is a continuation of Muddy Water, a student-led ArtsDoc Living Learning Community initiative. This initiative has already facilitated learning about sustainable coffee at the DKU-based bi-weekly farmer’s market and in the Kunshan-Suzhou area at Yuefeng Island Organic Farm and local cafes. Members of the Muddy Water initiative traveled to Thailand this summer as part of the course MEDIART 198 where they collaborated with local communities to document environmental and social change in the region through film and artistic work. Members became familiar with a range of local perspectives, learned about organic farming practices, and bought coffee from organic permaculture farms in the Huai Ma Kliang community. The ethically sourced coffee will be used for weekly, volunteer-based coffee workshops.

Muddy Water Coffee will function as a creative space where activities, such as film screening and ceramic events, will be held. It will be a platform for students to discuss pressing issues such as coffee consumption, environmental affairs, and sustainable farming. Members create sustainability models not just for coffee, but to elicit general consideration for ethically sourcing common commodities of daily life on the DKU campus and beyond. This research project will facilitate studying the ecosystems, communities, and commodity chains around coffee through an anthropological lens.

The application of ethnographic film and hands-on participation in coffee workshops provide a window into the lived experiences of individuals and communities. The faculty supervisor and existing student research team plan to return to the communities visited in Thailand in December and share their film work to ensure local people have agency in the way that their story is shared. Teaching and learning with ethnographic film engages students and audiences with ecological-historical-cultural relationships. Ethnographic film and art can illustrate anthropological concepts, offer understanding of fieldwork through the senses, and foster critical thinking about cultural representation and the filmmaking process itself. The methodical approach of returning to communities to ensure knowledge is co-created by local perspectives exposes students to ethical considerations, and fieldwork challenges that ethnographers face. The equipment used for filming includes analog cameras that require film processing in a dark room. Analogue is becoming a more popular medium of choice with the use of AI on the rise. The negatives made in analogue production create authenticity and observational evidence in a world that is drifting towards deep fakes through generative AI.

These types of AI production models not only allow people to make fake materials, but their uses limit face-to-face interactions. The coffee shop space on the other hand creates a space for face-to-face interactions to take place and foster connections while discussing themes relevant to students’ generation. By screening the film at DKU and creating other forms of media to convey that drinking coffee has implications for communities that are closer to us than they seem, the project intends to shed light on other social and environmental issues associated with this commodity chain, and spark discussions and ideas surrounding sustainability and alternative modes of consumption.

Humanoid robots have the potential to perform a wide range of tasks in human-centered environments, but enabling them to understand, plan, and execute complex real-world behaviors remains a major challenge. This project aims to develop intelligent humanoid robots by combining Vision-Language-Action (VLA) models with World Action Models (WAMs). VLA models provide the robot with the ability to understand visual observations and natural-language instructions and translate them into meaningful actions, while World Action Models learn predictive representations of how the physical world evolves under different actions. We will investigate how these two paradigms can be integrated to enable humanoid robots to perform complex, long-horizon tasks through closed-loop perception, prediction, planning, and action. Particular interests include whole-body manipulation, locomotion, human-robot interaction, learning from human demonstrations, and transferring knowledge from large-scale video and multimodal foundation models to real-world humanoid robots.