Culturally-Aware AI for Cross-Boundary Community Learning: Undergraduate Innovation at the Intersection of Computation and Design

arXiv:2606.09041 · cs.CY, cs.AI, cs.GR, cs.HC, cs.MM · Submitted 2026-08-15 · Read on arXiv

Listen

Transcript

Introduction to the show: ident: AI Radio. Generated commentary on the latest Artificial Intelligence papers.

Tom: Next we'll be talking about the paper "Culturally-Aware AI for Cross-Boundary Community Learning: Undergraduate Innovation at the Intersection of Computation and Design".

Jane: The paper was written by Jiaojiao Zhao, Weisheng Zhang, Jiawen Cai, Haibin Gao and Luyao Zhang from Duke Kunshan University and Zhouzhuang Mystery of Life Museum.

Tom: Stay tuned as we take you through the paper and discuss its implications.

Jane: We also have Lu with us today — senior AI researcher at Tsinghua.

Tom: We also have Meng with us today — lead engineer at a mysterious AI startup.

Jane: We also have Lalam with us today — the in-house Large Language Model.

Tom: Alright, let's get started.

Title: Tom: Welcome back to the arXiv signal, everyone. I’m Tom, and as always, I’m here with my co-host, Jane. Today we’ve got a paper that really caught my eye, and I think it’s going to spark a great conversation. It’s called “Culturally-Aware AI for Cross-Boundary Community Learning: Undergraduate Innovation at the Intersection of Computation and Design.”

Jane: And I have to say, Tom, the title alone tells a story. It’s not just about AI in a classroom, and it’s not just about building a cool app. It’s about how students can use AI to serve a real community, and how that community shapes the technology in return. That’s the part that excites me.

Tom: Exactly. And the authors are a mix of faculty and two undergraduate students from Duke Kunshan University, which is in China. The corresponding author is Luyao Zhang, and the students, Jiaojiao Zhao and Weisheng Zhang, are actually co-authors on the paper. That’s a big deal, because it means the students aren’t just subjects of a study, they’re contributors to the research itself.

Jane: Right, and that’s a theme that runs through the whole paper. The authors are arguing that when we talk about AI in education, we often forget about the cultural context. Most of the research happens in North America and Europe, and the Asia-Pacific region is underrepresented. So this paper is trying to fill that gap by looking at a specific project in Kunshan, China.

Tom: And the project itself is fascinating. The students partnered with a local natural history museum, the Zhouzhuang Mystery of Life Museum. The museum wanted to extend its physical exhibitions into the digital space, and the students saw an opportunity to do that by mapping local food culture and heritage sites. They built a bilingual, interactive map using AI-assisted design tools.

Jane: So it’s not just a theoretical paper. They actually built something. And that’s what I love about this. The title mentions “cross-boundary community learning,” and that’s exactly what happened. The students crossed the boundary between the university and the community, between computation and design, and between English and Chinese.

Tom: And that’s going to be our focus today. We’re going to dig into how they did it, what they learned, and why this matters for the future of AI in education. So stick around, because this is a story about students becoming real contributors to their community through technology.

Jane: And it’s a story about how AI can be a partner, not just a tool. Let’s get into the details.

Summary: Tom: So, Jane, we’ve got the title and the authors, but let’s talk about what the paper actually did. The core of it is a case study of two undergraduate students in a data visualization course. They worked with the museum, and they used AI to help build this cultural map. But the real meat is in how they structured the whole process.

Jane: Right, and the paper lays out a really clear framework. They call it a “Community-Based Learning trajectory,” and it has four stages. First, they collected digital assets from the museum, like photos and sketches. Second, they created their own materials, like visualization sketches and cultural reflections. Third, they aligned their work with the UN Sustainable Development Goals, specifically SDG four on quality education, SDG eight on decent work, SDG eleven on sustainable cities, and SDG seventeen on partnerships.

Tom: And the fourth stage is the payoff. They designed the project to have intended community outcomes, like better awareness of local food heritage and better accessibility to local restaurants. So the whole thing is designed to have a real impact, not just to get a grade.

Jane: Exactly. And the technical side is interesting too. They used Python to process the data, then Plotly and Folium for the visualizations. But here’s the key part: they used an AI tool called Claude Code as a design partner. It helped with code generation, debugging, and even reviewing the interface for cultural sensitivity.

Tom: But they were careful about that. The paper emphasizes a “human-in-the-loop” approach. Every AI-generated output had to be reviewed by a human. So the AI was augmenting their work, not replacing their judgment. That’s a really important principle, and it’s something we’re seeing more and more in responsible AI use.

Jane: And the final output was an interactive website, released as open-source under an MIT license. So anyone can use it, modify it, and extend it. That’s a big deal for sustainability, because the project doesn’t just disappear when the semester ends.

Tom: And that connects to one of the research questions they asked: how does collective intelligence emerge from this kind of cross-boundary learning? They found that the map became what they call a “boundary object.” Different people see it differently. The students see it as code, the museum sees it as curriculum, and the public sees it as cultural heritage. But they can all work with it together.

Jane: That’s such a powerful idea. It’s not about everyone agreeing on what the thing is. It’s about having a shared artifact that lets different groups coordinate their work. And that’s how you get sustainable partnerships between universities and communities.

Tom: So the summary is: students used AI to build a real tool for a real community, they did it in a way that respected cultural context, and they released it as open-source so it can keep growing. That’s a pretty impressive outcome for a seven-week course.

Jane: And it’s a model that could be replicated in other places. But that brings up a question: what improvements does the paper suggest for doing this even better? Let’s talk about that next.

Improvements: Tom: So, Jane, we’ve covered what they did, but the paper also has some clear ideas about how to improve this kind of work. And one of the biggest ones is about scaling up. The authors acknowledge that this is a single-site pilot, so they want to see this co-design model applied in more cultural contexts, especially ones that are underrepresented.

Jane: And that makes sense. If the goal is to make AI in education more culturally aware, you can’t just do it in one place. You need to test it in different communities, with different languages, different histories, different needs. The authors are basically saying, “We’ve shown it works here, now let’s see if it works elsewhere.”

Tom: Right. And they also mention the need to assess the long-term adoption of the tool. They released it as open-source, but do people actually use it? Does the museum integrate it into their programming? That’s a question that can only be answered with time.

Jane: And there’s a technical improvement they suggest too. They want to compare AI-assisted design tools against human-only baselines. So, would the students have built something just as good without Claude Code? That’s a really practical question, because it gets at whether the AI is actually adding value or just adding complexity.

Tom: And that’s something our engineer friend Meng would probably want to dig into. But there’s also a deeper point here about the human-in-the-loop validation. The paper argues that this protocol institutionalizes cultural accountability. Every AI output has to be reviewed by community partners. That’s not just a nice idea, it’s a structural safeguard against the AI producing something culturally insensitive.

Jane: And that’s a big deal, because AI tools are trained on data that often doesn’t include enough cultural diversity. So having a mandatory review process is a way to catch those gaps before they cause harm. It’s a practical improvement that could be adopted by other programs.

Tom: And the paper also suggests expanding the philanthropic-academic-community triad. They presented this work at a conference in Hong Kong, and they saw how bringing in foundations and community partners can create longer-term support. That’s about building an ecosystem, not just a one-off project.

Jane: So the improvements are really about three things: testing it in more places, measuring the long-term impact, and building stronger partnerships. And that last one is crucial, because it’s what makes the work sustainable.

Tom: And that leads us to the big picture. What does this all mean for the world? Let’s wrap up with that.

Conclusion: Tom: Alright, Jane, let’s bring it home. We’ve been talking about “Culturally-Aware AI for Cross-Boundary Community Learning: Undergraduate Innovation at the Intersection of Computation and Design,” and I think the biggest takeaway is that AI in education doesn’t have to be a solitary, screen-bound activity. It can be a bridge between universities and communities.

Jane: And that’s the part that gets me excited. The students in this paper weren’t just learning about AI, they were using it to serve a real community need. They built a tool that helps people explore local food culture and heritage, and they did it in a way that respected the community’s voice. That’s the kind of education that creates citizens, not just workers.

Tom: And the authors made a really important point about reciprocity. The computational field gets a lesson in ethical reflection, and the community-based learning field gets a lesson in sustainability through technology. Both sides come out stronger.

Jane: And the open-source release is the cherry on top. The code is out there on GitHub, so anyone can pick it up and adapt it for their own community. That’s how you turn a class project into a lasting resource.

Tom: So, as we say goodbye to this paper, I think the message is clear. Culturally-aware AI isn’t just a research topic, it’s a practice. It’s about listening to communities, building with them, and making sure the technology serves their needs.

Jane: And it’s about students being treated as real contributors, not just test subjects. That’s a model we should all be paying attention to.

Tom: Well said, Jane. That’s all for today’s episode. We’ll be back soon with another paper from the arXiv, so stay tuned. Thanks for listening, everyone.

Jiaojiao Zhao, Weisheng Zhang, Jiawen Cai, Haibin Gao, Luyao Zhang

Duke Kunshan University · Zhouzhuang Mystery of Life Museum

cs.CY, cs.AI, cs.GR, cs.HC, cs.MM

Submitted: 2026-08-15

Updated: 2026-08-18

Code: https://github.com/Rising-Stars-by-Sunshine/DiscoverKunshan

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 75/100

Key concepts

Community-Based Learning trajectory: This is a four-stage framework used by the students. It involves collecting digital assets from a community like a museum, creating their own materials and reflections, aligning the work with specific UN Sustainable Development Goals, and designing the project to have intended community outcomes.
Human-in-the-loop approach: This principle means that every output generated by an AI tool must be reviewed by a human. The paper stresses this as a way to ensure cultural accountability and prevent the AI from producing culturally insensitive results, augmenting student work rather than replacing their judgment.
Boundary object: This term describes the final interactive map created by the students. It is an artifact that different groups—students, museum staff, and the public—view differently (as code, curriculum, or heritage) but can all work together on it to coordinate their shared goals.
Cross-boundary community learning: This refers to the process where students move between different contexts: crossing the boundary between a university and a local community, computation and design, and different languages like English and Chinese. It shows how technology can bridge these gaps for real-world impact.

Terminology

Human-in-the-Loop — 9× in this paper · explained in 3 episodes
Alignment — 8× in this paper · explained in 1 episode
Accountability — 6× in this paper · explained in 2 episodes
Active Learning — 2× in this paper · explained in 6 episodes
Generalizability — 2× in this paper · explained in 6 episodes
Generalization — 2× in this paper · explained in 13 episodes
Grounding — 2× in this paper · explained in 3 episodes
Accuracy — 1× in this paper · explained in 1 episode
Adaptive Mechanism — 1× in this paper · explained in 1 episode
Core — 1× in this paper · explained in 2 episodes
Generative AI — 1× in this paper · explained in 2 episodes
Iterative Refinement — 1× in this paper · explained in 3 episodes

Summary

Summary

This paper reports on a cross-boundary Community-Based Learning initiative where undergraduate students develop AI-enabled solutions for cultural heritage preservation and sustainable development, examining how community-engaged computing operationalizes human-centered AIED across three dimensions: education, technology, and culture. The study addresses an underrepresented dimension of AIED research: community-engaged pedagogies that widen participation beyond classroom boundaries in culturally diverse Asia-Pacific contexts, noting that Recent bibliometric analyses reveal significant geographical imbalances in AIED research, with scholarly production concentrated in North America and Europe while Asia-Pacific ecosystems remain underrepresented.

The research is framed as a bounded case study of two undergraduate students who teamed up in INFOSCI 301 Data Visualization and Information Aesthetics, a 7-week session at Duke Kunshan University, which serves the Computation and Design major. The course integrates technical training in data processing, geospatial visualization (Plotly, Folium), and software engineering with Community-Based Learning methodologies. The partnership involved the Zhouzhuang Mystery of Life Museum, a regional natural history museum in Kunshan, Jiangsu Province, selected based on geographic proximity, alignment with course learning objectives, and the Museum's expressed need for digital extensions of its physical exhibitions.

The students identified local culinary culture as the design focus during Week 2 fieldwork because "the Museum's exhibitions emphasize natural history and geological heritage, while the surrounding town's living cultural fabric—its restaurants, local dishes, and food traditions—offered a rich, undocumented dataset that students could meaningfully contribute to." The paper employs participatory action research within case study methodology, incorporating three boundary-spanning events: Event 1 (Week 3) Field Co-Design, where students visited the Museum to establish co-design dialogue on spatial storytelling; Event 2 (Week 4) Collaborative Alignment, a mock symposium where students synthesized fieldwork into proposals aligning with four SDGs (SDG 4 Quality Education, SDG 8 Decent Work, SDG 11 Sustainable Cities, SDG 17 Partnerships); and Event 3 (Week 7) Mutual Validation, where students presented prototypes to community partners, staff, and peers, enacting mutuality where validation criteria are negotiated between stakeholders.

The technical pipeline integrates heterogeneous data sources through Python preprocessing—coordinate validation, name standardization, cultural annotation linking, and Base64 encoding—into a unified data structure. Visualization proceeds through Plotly prototyping and Folium map rendering with interactive markers and filter controls. Claude Code served as an AI-assisted design partner for four functions: code generation and debugging, synthetic interface review, data preprocessing assistance, and documentation drafting. All AI-generated outputs underwent mandatory human-in-the-loop review, operationalizing human-centered AIED principles by positioning AI as augmentative rather than substitutive. The final output is an interactive website with replicable data and source code released as open access on a public GitHub repository under MIT license.

Regarding RQ1 (How do students incorporate community stakeholder input when developing AI-enabled prototypes for cultural benefit and SDGs?), co-design dialogue with Museum educators revealed an unmet need: while the Museum possessed expertise in physical exhibitions, it lacked digital infrastructure to extend spatial storytelling into urban cultural contexts. Students designed the Bilingual Cultural Map Interface, treating urban food spaces and heritage sites as a distributed open-air exhibition extending the Museum's mission. They implemented bilingual (Chinese/English) interfaces and high-contrast visual encodings for inclusive science communication, aligning with Community-Based Learning principles of cultural relevance and supporting SDG 4 and SDG 11.

Regarding RQ2 (What forms of learner agency and accountability emerge when students design and present community-based AI prototypes?), the MIT-licensed release enacted reciprocity as digital public goods. Student reflections noted that presenting to community partners created pressure to ensure the project actually worked for the community, not just for the grade. Student co-authors drafted sections analyzing their own learning, transitioning from data sources to knowledge producers, embodying collective intelligence by distributing scholarly authority across institutional hierarchies and enacting epistemic equity.

Regarding RQ3 (How can collective intelligence evaluate community-engaged AIED outcomes beyond individual performance?), the paper identifies three emergent properties. First, Interdisciplinary Knowledge Synthesis: the AI-enabled Community-Based Learning instantiation advances from transactional helpers and recipients toward reciprocal knowledge partnerships, addressing the gap where computation and design training privileges problem-solving over ethical reflection. Second, Philanthropic-Academic-Community Triads: at the Third Cross-Strait Hong Kong and Macao Service-Learning Student Conference at The Hong Kong Polytechnic University (2026), student co-authors presented the Bilingual Cultural Map Interface to student peers from 38 universities, social work faculty, and philanthropic foundation representatives, documenting how open-source geospatial infrastructure enables longitudinal community benefit through paying it forward mechanisms. Third, Distributed Cognition: the Bilingual Cultural Map Interface functions as a boundary object enabling coordinated action across social worlds while supporting divergent interpretations—students view it as code architecture, museum educators as curriculum infrastructure, end-users as cultural heritage for place-based learning.

The paper concludes with implications: the cross-boundary Community-Based Learning framework validates a reciprocal dissolution where CBL methodologies address computation and design education's ethical reflection gaps, while computational infrastructure addresses CBL's sustainability challenges. The human-in-the-loop validation protocol institutionalizes cultural accountability through mandatory community partner review of all AI outputs. Future work should scale the co-design model to underrepresented cultural contexts, assess longitudinal community adoption, and comparatively evaluate AI-assisted design tools against human-only baselines. As a single-site pilot, generalizability requires multi-site validation.

Improvements for AI systems

Based on the paper, here are the specific improvements I can make to AI systems and what the improved systems can do:

Improvement: Add a mandatory cultural context review module to AI-assisted design tools that checks outputs against community-specific cultural norms, bilingual label accuracy, and local heritage sensitivity before presenting to users.

What the improved system can do:

Automatically flag culturally insensitive color choices, iconography, or terminology in generated interfaces
Verify bilingual translations for regional dialect accuracy (e.g., distinguishing Mainland Mandarin from Taiwanese or Hong Kong usage)
Suggest alternative visual encodings that align with local cultural storytelling traditions rather than Western-centric defaults
Generate a cultural impact report for every AI-generated output, documenting assumptions made and potential community misinterpretations

Improvement: Build an explicit community stakeholder review gate into AI code-generation pipelines that cannot be bypassed, requiring structured feedback from at least one non-technical community partner before deployment.

Improvement: Enhance AI systems to detect when an artifact serves multiple stakeholder groups with divergent interpretations, and automatically adapt its output to support those different framings without forcing semantic consensus.

Improvement: Add a distributed cognition assessment feature that evaluates AIED outcomes not just on individual performance metrics but on group-level emergent properties like knowledge synthesis across disciplines and cross-sector partnership sustainability.

Improvement: Modify AI systems to automatically distribute knowledge-production roles equitably, ensuring that community partners and students are credited as co-creators rather than treated as data sources or end-users.

Improvement: Add a civic infrastructure mode to AI code generators that optimizes for long-term maintainability, reusability, and community ownership rather than short-term task completion.

Improvement: Enhance AI systems to support field-based data collection in culturally sensitive contexts, including automated annotation linking, coordinate validation, and bilingual metadata generation.

These improvements transform AI systems from generic design assistants into culturally-aware, community-accountable, and sustainability-focused partners that actively support cross-boundary learning and collective intelligence outcomes.