Co-Creating a Regional Sustainability Hub: Conversational AI, Community Engagement, and Local Data for Computing in Place Daniel Pittman Department of Computer Sciences Metropolitan State University of Denver Denver, Colorado, USA dpittma8@msudenver.edu Alyssa Williams Department of Computer Sciences Metropolitan State University of Denver Denver, Colorado, USA awill157@msudenver.edu Kerstin Haring Department of Computer Science University of Denver Denver, Colorado, USA kerstin.haring@du.edu Jessica Salo Department of Geography, GIS, and Sustainability University of Northern Colorado Greeley, Colorado, USA jessica.salo@unco.edu Gregory Newman CitSci Colorado State University Fort Collins, Colorado, USA gregory.newman@colostate.edu Alexis Kennedy Department of Political Science Colorado State University Fort Collins, Colorado, USA alexis.kennedy@colostate.edu Sarah Newman CitSci Colorado State University Fort Collins, Colorado, USA sarah.newman@colostate.edu Sylvester Kalevela School of Engineering Colorado State University Pueblo Pueblo, Colorado, USA sylvester.kalevela@csupueblo.edu Abstract The Sustainability Hub is a community-driven digital infrastructure project that connects policymakers, researchers, and community members across Colorado through a shared platform for sustain- ability and well-being data. Grounded in a place-based computing approach, the Hub integrates conversational AI, open-source tools, and community asset mapping to support regional decision-mak- ing and resilience. This paper presents early-stage development and future directions. We introduce Bili, a natural language assis- tant that lowers technical barriers to data access and BiliCore, a framework for evaluating large language models in sustainability contexts. Another key component is the implementation of survey- informed Communities of Interest to support collaboration and engagement. We describe our co-design methodology, beta testing strategy, and participatory outreach efforts, offering a real-world case of computing in place that bridges technical innovation with social infrastructure. CCS Concepts • Computing methodologies → Machine learning; Natural language processing; • Human-centered computing→ Human Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for third-party components of this work must be honored. For all other uses, contact the owner/author(s). COMPASS ’25, Toronto, ON, Canada © 2025 Copyright held by the owner/author(s). ACM ISBN 979-8-4007-1484-9/25/07 https://doi.org/10.1145/3715335.3736321 computer interaction (HCI); • Applied computing → Com- puting in government; • Information systems → Geographic information systems; Open source software. Keywords sustainability, computing in place, community engagement, partici- patory design, conversational AI, large language models, retrieval- augmented generation, human-computer interaction, natural lan- guage processing, geographic information systems ACM Reference Format: Daniel Pittman, Alyssa Williams, Kerstin Haring, Jessica Salo, Gregory Newman, Alexis Kennedy, Sarah Newman, and Sylvester Kalevela. 2025. Co-Creating a Regional Sustainability Hub: Conversational AI, Commu- nity Engagement, and Local Data for Computing in Place. In ACM SIG- CAS/SIGCHI Conference on Computing and Sustainable Societies (COMPASS ’25), July 22–25, 2025, Toronto, ON, Canada. ACM, New York, NY, USA, 6 pages. https://doi.org/10.1145/3715335.3736321 1 Introduction As computing technologies become more embedded in social, en- vironmental, and political contexts, sustainability research must consider how digital systems are governed and aligned with com- munity needs. These questions are especially urgent as machine learning, data platforms, and automation are rapidly integrated into public-sector decision-making and environmental monitoring. Emerging work across geospatial AI, sustainability science, and citizen engagement has underscored the importance of designing computing systems that are attuned to place, enabling context- aware decision-making and grounded collaboration [18, 21, 26]. This late breaking work presents the early stage development of the Sustainability Hub, a regional digital infrastructure initiative in 746 https://orcid.org/0000-0002-0888-3587 https://orcid.org/0009-0004-3965-3647 https://orcid.org/0000-0001-6338-5575 https://orcid.org/0000-0003-0516-4189 https://orcid.org/0000-0003-0503-5782 https://orcid.org/0000-0002-6637-8976 https://orcid.org/0009-0007-3094-9662 https://orcid.org/0009-0006-9903-4367 https://doi.org/10.1145/3715335.3736321 https://doi.org/10.1145/3715335.3736321 http://crossmark.crossref.org/dialog/?doi=10.1145%2F3715335.3736321&domain=pdf&date_stamp=2025-07-21 COMPASS ’25, July 22–25, 2025, Toronto, ON, Canada Pittman et al. Colorado. Our goal is to create a platform that bridges fragmented sustainability and well-being data while making information more accessible and actionable for three core groups: community mem- bers, researchers, and policymakers. Unlike tools that prioritize abstract modeling or generic policy indicators, the Hub is explicitly place-based and community-informed. It supports the discovery and use of data by those facing sustainability challenges. The Sustainability Hub emerged from the convergence of local needs, technological advances, and institutional collaboration. In Colorado, rapid development and climate variability have exposed the limits of traditional data systems, particularly for rural commu- nities and under-resourcedmunicipalities. Advances in open-source tools, geospatial infrastructure, and natural language processing now offer opportunities to design more accessible, community- centered platforms. In response, we are designing a modular system to address three challenges: fragmented local data, technical barriers to interpreta- tion, and the lack of inclusive civic tools. Key components include Bili, a natural language interface for querying sustainability data, BiliCore [29], an evaluation framework for sustainability-focused LLMs, and Communities of Interest, which enables users to organize around place-or issue-based concerns. 2 Background and Motivation Colorado is undergoing transformation driven by environmental, social, and technological pressures. Urbanization, climate variability, water scarcity, and demographic shifts have heightened the need for tools that support regionally grounded sustainability planning. Yet efforts to address these challenges remain fragmented across disconnected datasets, disciplinary silos, and uneven technological capacity. Studies show that civic data platforms often struggle with equitable access, particularly in rural or under-served areas [7, 27]. The Sustainability Hub responds by offering a unifying digital infrastructure grounded in community participation and AI-en- hanced data accessibility. Drawing from geospatial AI and open data platform research [18, 21, 28], we emphasize that sustainabil- ity is inherently local and that data infrastructures must reflect regional context. The Hub explores Large Language Models (LLMs), Retrieval- Augmented Generation (RAG), and geospatial indexing as methods for data democratization. This work draws from Human-Computer Interaction (HCI) and computing for sustainability, particularly approaches that examine the social and political dimensions of data systems. Scholars have shown that platforms are not neutral; they encode assumptions about what counts as knowledge, who gets to produce it, and how it is used [30, 34]. Technologies must be evaluated for both their function and the values and hierarchies they reinforce. As sustainability experts argue, digital systems must account for their material and environmental footprints, not just the outcomes they support [24]. In the context of AI and sustainability, systems that privilege institutional data or obscure provenance can unintentionally reproduce power imbalances and limit local relevance [27]. To resist this trend, we draw on “interface justice” and partici- patory design. Our co-design strategies, rooted in citizen science and public engagement literature [11, 13, 20], aim to position stake- holders as active co-authors of the platform. Rather than treating stakeholders as end-users, we build on participatory science models [11] that emphasize iterative co-creation and long-term steward- ship. This ethos informs key design principles: • Universal Usability: Interfaces for non-experts, including mobile-first layouts and multilingual support. • Contextual Insight: Metadata, provenance, and explana- tion to support interpretation. • ParticipatoryAuthorship: Community-driven submission, annotation, and curation of datasets. Figure 1: Bili chatbot prototype showing text-based responses to sustainability-related user queries In parallel with community outreach, we are building technical infrastructure to support AI-powered research. This includes the Bili conversational agent and BiliCore [9], an open framework for evaluating LLMs in sustainability tasks using cloud and local mod- els [31, 33, 35]. As shown in Figure 1, our goal is not just accuracy, but usability, transparency, and trust. Following critiques of civic AI governance and critical HCI [2, 31], we view LLMs not as au- tonomous decision-makers but as scaffolds for traceable, contextual inquiry. By embedding these tools within regional governance and geospatial data systems, the Hub operates at the intersection of civic needs and AI innovation, supporting sustainable futures that are co-created, accountable, and grounded in place. 3 Community Asset Mapping and Co-Design Process In order to ensure that our planned in-person community engage- ment sessions are as meaningful and inclusive as possible, we be- gan by cataloging organizations and communities across Colorado engaged with sustainability efforts across the three pillars of sus- tainability: Economic, Ecological, and Societal. This resulted in a catalog of more than 1,200 entities working in land and water man- agement, environmental education, public health, and food systems. Data sources included CSU Extension consultation, public datasets, and community-science partnerships via CitSci.org. An asset map, shown in Figure 2, was created to build a relational view of Colorado’s sustainability infrastructure. Organizations were tagged by domain, geography, affiliation (e.g., tribal, municipal, 747 Computing in Place with the Sustainability Hub COMPASS ’25, July 22–25, 2025, Toronto, ON, Canada Figure 2: Community asset locations across Colorado from the initial mapping phase nonprofit), and data holdings. The process revealed interoperability gaps, especially in rural areas, and highlighted clusters-such as watershed coalitions and air quality networks-that would benefit from better data-sharing tools. To understand how these groups engage with data, we piloted a statewide survey with questions on data types, tools (e.g., Ar- cGIS, spreadsheets), staffing, sharing challenges, and infrastructure needs. We also explored attitudes toward AI and natural language interfaces to guide Bili’s development. Distributed through regional partners, email campaigns, and planning meetings, the survey re- ceived over 130 responses. Preliminary findings include: • Most organizations collect data, but lack the capacity to analyze it meaningfully. • The demand for simplified discovery tools is high, especially with spatial filtering. • Participants value tagging, curation, and storytelling, high- lighting metadata and narrative framing. • There is cautious optimism toward AI, contingent on trans- parency and user control. These findings confirm the need for low-barrier interfaces, par- ticipatory workflows, and shared infrastructure to connect siloed efforts, consistent with observed patterns in participatory science research [11, 16, 23]. They are directly informing the Hub’s beta prototype, guiding features like regional filters, data views, and early Communities of Interest. To keep the Hub grounded in user needs, we are adopting a phased co-design process. In Summer 2025 we will employ field technicians to conduct in-person interviews and canvassing ses- sions to gather community feedback. In fall 2025, we will host listening sessions and co-design workshops with a cross-section of organizations identified in the asset mapping, those who par- ticipated in the survey, and community members with whom we connected in summer. These sessions will span urban, rural, and frontier contexts. Participants will include representatives of governments, non- profits, academic institutions, tribal groups, and local residents. In partnership with CSU Extension, sessions will use accessible language, hands-on exploration, and collaborative critique to: • Identify regional priorities and data gaps. • Evaluate Bili and CoI usability. • Surface trust and representation concerns. • Invite contributions of locally relevant data. Workshops will be documented through field notes, consented recordings, and participatory rubrics. Feedback will be synthesized and used to refine the design. Drawing on participatory design practices [11, 13, 20], this process treats iteration not as user testing, but as civic capacity-building, strengthening relationships, data confidence, and local voice. This phase extends our survey work into sustained partnerships. It will allow us to engage directly with organizations, build trust, and gather critical feedback to inform user interface design, data presentation and visualization approaches, and the overall structure of the platform. This reinforces our commitment to principles of HCI and community-driven design, ensuring that future iterations of the Hub are shaped by the real-world goals, constraints, and interpretive needs of those it is meant to serve. 4 Bili and BiliCore: A HCI-Informed Conversational AI Framework A central feature of the Sustainability Hub is Bili, a conversational AI assistant designed to reduce technical barriers to accessing com- plex sustainability data. It connects users, ranging from rural plan- ners to researchers, with structured and unstructured data through natural language. Bili is grounded in Sustainable HCI, Participatory Design, and Critical HCI principles, which examine the social, epistemic, and en- vironmental dimensions of digital systems [1, 8, 22]. Our approach, rooted in citizen science and civic informatics [13, 20], recognizes the need for participatory infrastructure attentive to power, repre- sentation, and inclusion. We launched early interface prototyping and public-facing de- velopment through two HCI courses at the University of Denver. Student designs, shown in Figure 1 and Figure 3, were grounded in HCI best practices and inform wireframes and use cases for co- design workshops. A public website shares updates, user stories, and will phase in access to Bili and the Communities of Interest platform in 2025. We are critically attentive to how language models may repro- duce institutional biases or exclude underrepresented sources.Many community organizations hold valuable sustainability knowledge that lacks public infrastructure. The Hub will support direct data contributions, source transparency, contributor attribution, and sig- naling of uncertainty or disagreement, enabling diverse knowledge formats in an inclusive environment. Bili combines custom retrieval pipelines with LLMs from AWS Bedrock, OpenAI, Google Cloud Platform, and locally hosted mod- els. It is embedded in the Hub’s full-stack MERN web application, MongoDB [25], Express.js [4], React [32], Node.js [10], with Post- GIS [3] supporting geospatial data and APIs enabling querying, visualization, and summarization. Role-based views accommodate different user needs. Bili translates plain-language queries into interpretable outputs, combining summaries, citations, and dynamic maps. For example, a user might ask, “What are the top drought resilience initiatives in 748 COMPASS ’25, July 22–25, 2025, Toronto, ON, Canada Pittman et al. Figure 3: Bili chatbot prototype interface with textual sum- marization and geospatial visualization of sustainability data in response to user query. western Colorado?” and receive (1) a summary, (2) linked sources, and (3) geospatial data visualizations. Figure 4: Architecture diagram of Bili + BiliCore stack with user flow Designing this interface involves trade-offs between precision, transparency, and accessibility. Figure 4 shows how we apply best practices in civic LLM design [12, 14, 26] through: • Interactive Query Interface: Threaded conversations with memory support for sustainability topics • Document-Aware Responses: LangGraph and LangChain agents ground outputs in local reports and geospatial datasets • Visual Summarization: Maps, graphs, and timelines con- textualize retrieved results • Editable History and Feedback: Users view, edit, and rate responses; feedback informs improvement To evaluate and guide these systems, we developed BiliCore [29], an open-source framework for testing LLMs in sustainability con- texts. Generic benchmarks often miss the complexity of sustainabil- ity queries, which require geospatial reasoning, provenance track- ing, and synthesis across formats [15, 33, 35]. Built with LangChain [5], LangGraph [6], and Streamlit [19], BiliCore supports: • Cross-Model Evaluation: Compare GPT-4, Gemini, Nova, LLaMA, and others for consistency and relevance • Custom RAG Pipelines: Dynamically configure document stores (e.g., FAISS, OpenSearch), retrievers, and prompts • Interactive Testing: Preserve and re-run conversations to assess reproducibility and robustness • Configurable Parameters: Tune behaviors across task types and model switches We plan to use BiliCore to stress-test prompts, explore failure modes, and analyze behavior on questions from community part- ners such as hallucination rates and spatial reasoning accuracy. The framework is publicly available on GitHub1, and we are de- veloping support for community-submitted prompts and shared benchmarks. BiliCore also informs ethical and technical design decisions in Bili[2, 17], including prompt engineering, fallback logic, and inter- face strategy (e.g., disambiguating counties from municipalities). In doing so, it contributes to civic AI governance and HCI-aligned evaluation practices [24, 27, 31, 34]. 5 Future Work In summer 2025, we will launch a public beta of the Sustainability Hub, transitioning from prototyping to real-world testing. This phase will assess how diverse users engage with sustainability data in place-based, socially meaningful contexts, with a focus on civic and collaborative use. A central feature of the beta is the introduction of Communities of Interest (CoIs), user-formed groups organized around specific topics, locations, or data sets. CoIs serve as social and epistemic infrastructure, enabling users to explore data through shared con- cerns and lived experience (e.g., fire recovery in Boulder County or water equity in the Arkansas River basin). Each CoI space will support: • User Profiles and Affiliation Tags: Members identify roles (e.g., nonprofit staff, policymaker) and affiliations for context • Story-Based Data Exploration: Users curate “storylines” linking datasets and annotations to local issues • Collaborative Annotation: Community members flag is- sues, suggest metadata, and contextualize datasets • Discussion and Resource Sharing: Threaded comments and shared bookmarks support sustained dialogue This model builds on participatory infrastructure literature [16, 23, 34] and survey results indicating that many organizations are active stewards, not passive consumers, of local knowledge. CoIs 1https://github.com/msu-denver/bili-core 749 https://github.com/msu-denver/bili-core Computing in Place with the Sustainability Hub COMPASS ’25, July 22–25, 2025, Toronto, ON, Canada reflect a core principle: data becomes actionable when embedded in relational, community-based contexts. Through annotation, remix- ing, and shared curation, the Hub supports inquiry that foregrounds infrastructure, power, and access. We will evaluate CoIs using a mix of analysis, interviews, and participatory methods. Rather than a fixed feature, CoIs function as a design hypothesis: structured data spaces can scaffold collabo- ration, trust, and civic learning. Human-in-the-loop evaluation will also advance through Bil- iCore. Planned features include: • Latency tracking, token usage visualization, and prompt–response ratio tools • A user rating interface for assessing relevance, completeness, accuracy, and tone • Role-specific rubrics for policymakers, researchers, and com- munity members • Task-type adaptations for summarization, explanation, and comparison This builds on recent work in inclusive, domain-specific LLM evaluation [2, 24] and will be piloted during co-design workshops. To grow our user and contributor base, we will pursue the fol- lowing outreach strategies: • Webinars and Demos: Co-hosted with regional partners to present features and gather early feedback • Conferences and Publications: Engagement at AGU, Esri UC, CHI, AAAI, SIGSPATIAL, and similar venues • CommunityDocumentation: Support for tutorials, guides, and data stories grounded in regional use • Contributor Training: Asynchronous onboarding, office hours, and technical support for collaborators These activities aim to foster not only adoption, but a contrib- utor ecosystem where local organizations, practitioners, and res- idents shape and evolve the platform. As a living infrastructure, the Hub will adapt alongside community needs, technical progress, and participatory governance to support inclusive, actionable, and grounded sustainability systems. 6 Conclusion The Sustainability Hub presents a grounded, multi-layered ap- proach to regional AI infrastructure. Unlike generic civic dash- boards or standalone tools, it is designed around Colorado’s distinct geographic, institutional, and social contexts. By integrating local data systems, participatory design, and open-source AI, the Hub be- comes a flexible system that evolves alongside regional challenges and civic priorities. This work addresses several key design questions relevant to sustainability and public-sector computing: • Whose data is visible? Smaller municipalities, conserva- tion groups, and nonprofits often maintain valuable datasets disconnected from broader systems. The Hub supports par- ticipatory authorship and collaborative curation to elevate these perspectives. • How are systems shaped by place? Regional divides, en- vironmental concerns, and local policies influence data se- lection, interface behavior, and conversational framing. • Can LLMs support civic reasoning? Bili and BiliCore are designed to explore this question, emphasizing grounding in real data, transparent sourcing, and human-in-the-loop feedback. The Hub functions as a hybrid infrastructure: part data com- mons, part AI sandbox, part community platform. Natural language interfaces lower barriers while preserving local context and agency. LLMs are not framed as decision-makers, but as tools for situated inquiry and collaborative exploration. Our HCI-informed, community-led design process remains cen- tral to this vision. Through completed cataloging and survey efforts, as well as upcoming canvassing sessions and co-design workshops, we aim to build a system aligned with real-world goals, constraints, and interpretive needs. Looking ahead, we see the Hub as a living infrastructure. It is an adaptable, open-source model for civic technology that is regionally responsive and broadly extensible. With a public beta scheduled for summer 2025, we welcome collaboration to advance place-aware, participatory approaches to sustainability computing. Acknowledgments This material is based on work supported by the National Science Foundation under Grant No. 2318730 and the NAIRR Pilot under Award No. NAIRR240197. We thank the undergraduate research assistants at MSU Denver, Colorado State University Fort Collins and Pueblo who contributed to the data visualizations and graphics in this article, including the following: Gabriel Castro, Angie Cruz, Maria Bhatti, Nyla Bickham, Whitney Haddad, Vashti Trujillo, and Cindy Vo. We acknowledge the use of generative AI tools, including Chat- GPT, for iterative text development and outline generation. The final content was reviewed and revised by the authors. References [1] Eli Blevis. 2007. Sustainable Interaction Design: Invention & Disposal, Renewal & Reuse. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI ’07). 503–512. doi:10.1145/1240624.1240705 [2] Yupeng Chang, Xu Wang, Jindong Wang, Yuan Wu, Linyi Yang, Kaijie Zhu, Hao Chen, Xiaoyuan Yi, Cunxiang Wang, Yidong Wang, Wei Ye, Yue Zhang, Yi Chang, Philip S. Yu, Qiang Yang, and Xing Xie. 2024. A Survey on Evaluation of Large Language Models. 15, 3 (2024), 39:1–39:45. doi:10.1145/3641289 [3] PostGIS Project Steering Committee. 2024. PostGIS: Spatial and Geographic Objects for PostgreSQL. https://postgis.net. Accessed March 22, 2025. [4] Express.js Contributors. 2024. Express - Fast, unopinionated, minimalist web framework for Node.js. https://expressjs.com. Accessed March 22, 2025. [5] LangChain Contributors. 2023. LangChain: Building Applications with LLMs through Composability. https://github.com/langchain-ai/langchain. GitHub repository. Accessed March 22, 2025. [6] LangGraph Contributors. 2024. LangGraph: Composable Multi-Actor RAG and AgentWorkflows. https://github.com/langchain-ai/langgraph. GitHub repository. Accessed March 22, 2025. [7] Oliver Dalby, Isadora Sinha, Richard K. F. Unsworth, Len J. McKenzie, Benjamin L. Jones, and Leanne C. Cullen-Unsworth. 2021. Citizen Science Driven Big Data Collection Requires Improved and Inclusive Societal Engagement. 8 (2021). https://www.frontiersin.org/articles/10.3389/fmars.2021.610397 [8] Paul Dourish. 2007. Responsibilities and Implications: Further Thoughts on Ethnography and Design. In Proceedings of the 2007 Conference on Designing for User eXperiences (DUX ’07). 1–14. doi:10.1145/3279974.3279981 [9] Luke Farchione and Daniel Pittman. 2025. Built for Sustainability, Designed for Versatility: A Customizable Framework for Evaluating Large Language Models Across Domains. NAIRR Pilot Inaugural Annual Meeting (2025). doi:10.5281/ ZENODO.15114563 [10] OpenJS Foundation. 2024. Node.js: JavaScript runtime built on Chrome’s V8 JavaScript engine. https://nodejs.org. Accessed March 22, 2025. 750 https://doi.org/10.1145/1240624.1240705 https://doi.org/10.1145/3641289 https://postgis.net https://expressjs.com https://github.com/langchain-ai/langchain https://github.com/langchain-ai/langgraph https://www.frontiersin.org/articles/10.3389/fmars.2021.610397 https://doi.org/10.1145/3279974.3279981 https://doi.org/10.5281/ZENODO.15114563 https://doi.org/10.5281/ZENODO.15114563 https://nodejs.org COMPASS ’25, July 22–25, 2025, Toronto, ON, Canada Pittman et al. [11] T. Frensley, Alycia Crall, Marc Stern, Rebecca Jordan, Steven Gray, Michelle Prysby, Greg Newman, Cindy Hmelo-Silver, David Mellor, and Joey Huang. 2017. Bridging the Benefits of Online and Community Supported Citizen Science: A Case Study on Motivation and Retention with Conservation-Oriented Volunteers. 2, 1 (2017), 1–14. doi:10.5334/cstp.84 Section: 1. [12] Julian “Iñaki” Goñi, Claudio Fuentes, andMaria Paz Raveau. 2023. An experiential account of a large-scale interdisciplinary data analysis of public engagement. 38, 2 (2023), 581–593. doi:10.1007/s00146-022-01457-4 [13] Gregory Newman, Newman Sarah, Scarpino Russell, Kaplan Nicole, Crall Alycia, and Lynn Stacy. 2020. Data Management and Visualization, Chapter 10. In Handbook of Citizen Science in Ecology and Conservation. University of California Press. [14] Yuanyuan Guo and Peng Dong. 2024. Factors Influencing User Favorability of Government Chatbots on Digital Government Interaction Platforms across Different Scenarios. 19, 2 (2024), 818–845. doi:10.3390/jtaer19020043 [15] Sungwon Han, Donghyun Ahn, Seungeon Lee, Minhyuk Song, Sungwon Park, Sangyoon Park, Jihee Kim, and Meeyoung Cha. 2024. GeoSEE: Regional Socio- Economic Estimation With a Large Language Model. doi:10.48550/arXiv.2406. 09799 arXiv:2406.09799 [16] Jitka Stilund Hansen, Signe Gadegaard, Karsten Kryger Hansen, Asger Væring Larsen, Søren Møller, Gertrud Stougård Thomsen, and Katrine Flindt Holmstrand. 2021. Research Data Management Challenges in Citizen Science Projects and Recommendations for Library Support Services. A Scoping Review and Case Study. 20, 1 (2021), 25. doi:10.5334/dsj-2021-025 [17] Ellen Hohma and Christoph Lütge. 2023. From Trustworthy Principles to a Trust- worthy Development Process: The Need and Elements of Trusted Development of AI Systems. 4, 4 (2023), 904–926. doi:10.3390/ai4040046 [18] Rania Elsayed Ibrahim, Islam Abou El-Magd, Tshiamo Motshegwa, Hamed Ab- delreheem Ead, Madara Ogot, Joseph Wafula, Abotalib Z. Abotalib, Nokuthula P. Mchunu, Sepo Hachigonta, John Kinyuru, and Elham M. Ali. 2024. The Role of Open Science and Geoinformatics in Advancing Sustainable Development Goals in Africa: A Strategic Framework and an Action Plan. 23 (2024), 47. doi:10.5334/dsj-2024-047 [19] Streamlit Inc. 2020. Streamlit: The fastest way to build and share data apps. https://github.com/streamlit/streamlit. GitHub repository. Accessed March 22, 2025. [20] Alan Irwin. 1995. Citizen Science: A Study of People, Expertise and Sustainable Development, by Alan Irwin. London: Routledge, 1995, 216 pp. Routledge. [21] Krzysztof Janowicz, Song Gao, Grant McKenzie, Yingjie Hu, and Budhen- dra Bhaduri. 2020. GeoAI: spatially explicit artificial intelligence tech- niques for geographic knowledge discovery and beyond. 34, 4 (2020), 625– 636. doi:10.1080/13658816.2019.1684500 Publisher: Taylor & Francis _eprint: https://doi.org/10.1080/13658816.2019.1684500. [22] Bran Knowles, Oliver Bates, and Maria Håkansson. 2018. This Changes Sus- tainable HCI. In Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems (CHI ’18). 1–12. doi:10.1145/3173574.3174045 [23] S. J. Lynn, N. Kaplan, S. Newman, R. Scarpino, and G. Newman. 2019. Designing a Platform for Ethical Citizen Science: A Case Study of CitSci.org. Citizen Science: Theory and Practice, 4(1), 14. 4, 1 (2019), 14. doi:10.5334/cstp.227 [24] Martinus Maslim, Hei-Chia Wang, Cendra Devayana Putra, and Yulius Denny Prabowo. 2024. A Trustworthy Automated Short-Answer Scoring System Using a New Dataset and Hybrid Transfer Learning Method. 8, 5 (2024), 37. doi:10. 9781/ijimai.2024.02.003 [25] Inc. MongoDB. 2024. MongoDB: The Developer Data Platform. https://www. mongodb.com. Accessed March 22, 2025. [26] Ehsan Nabavi, Holger R. Maier, Saman Razavi, Adrian Hindes, Mark Howden, Will Grant, and Sujatha Raman. [n. d.]. Potential Benefits and Dangers of Using Large Language Models for Advancing Sustainability Science and Communica- tion. ([n. d.]). https://www.authorea.com/doi/full/10.22541/essoar.171136837. 71755629?commit=b0e4bff51da5dc2ec8f3e2e77ebc2a7427d04d1f [27] Margaret O’Brien, Ruth Duerr, Riley Taitingfong, AndrewMartinez, Lourdes Vera, Lydia L. Jennings, Robert R. Downs, Erin Antognoli, Talya Ten Brink, Nicole B. Halmai, Dominique David-Chavez, Stephanie Russo Carroll, Maui Hudson, and Pier Luigi Buttigieg. 2024. Earth Science Data Repositories: Implementing the CARE Principles. 23 (2024), 37. doi:10.5334/dsj-2024-037 [28] Roberto Pierdicca and Marina Paolanti. 2022. GeoAI: a review of artificial intelli- gence approaches for the interpretation of complex geomatics data. 11, 1 (2022), 195–218. doi:10.5194/gi-11-195-2022 Publisher: Copernicus GmbH. [29] Daniel Pittman and Sustainability Hub Team. 2025. BiliCore: An Open Framework for Evaluating LLMs in Sustainability Contexts. https://github.com/msu-denver/ bili-core. Version 0.3.1. GitHub repository. Accessed March 22, 2025. [30] Eryn Rigley, Adriane Chapman, Christine Evers, and Will McNeill. 2023. An- thropocentrism and Environmental Wellbeing in AI Ethics Standards: A Scoping Review and Discussion. 4, 4 (2023), 844–874. doi:10.3390/ai4040043 [31] Tobias Schimanski, Andrin Reding, Nico Reding, Julia Bingler, Mathias Kraus, and Markus Leippold. 2023. Bridging the Gap in ESGMeasurement: Using NLP to Quantify Environmental, Social, and Governance Communication. doi:10.2139/ ssrn.4622514 [32] Meta Open Source. 2024. React: A JavaScript library for building user interfaces. https://react.dev. Accessed March 22, 2025. [33] Alexey Tikhonov. 2024. PLUGH: A Benchmark for Spatial Understanding and Reasoning in Large Language Models. (2024). doi:10.48550/ARXIV.2408.04648 Publisher: arXiv Version Number: 1. [34] Danilo Urzedo, Zarrin Tasnim Sworna, Andrew J. Hoskins, and Cathy J. Robinson. 2024. AI chatbots contribute to global conservation injustices. 11, 1 (2024), 204. doi:10.1057/s41599-024-02720-3 [35] Liuchang Xu, Shuo Zhao, Qingming Lin, Luyao Chen, Qianqian Luo, Sensen Wu, Xinyue Ye, Hailin Feng, and Zhenhong Du. 2024. Evaluating Large Language Models on Spatial Tasks: A Multi-Task Benchmarking Study. doi:10.48550/arXiv. 2408.14438 arXiv:2408.14438 [cs] 751 https://doi.org/10.5334/cstp.84 https://doi.org/10.1007/s00146-022-01457-4 https://doi.org/10.3390/jtaer19020043 https://doi.org/10.48550/arXiv.2406.09799 https://doi.org/10.48550/arXiv.2406.09799 https://arxiv.org/abs/2406.09799 https://doi.org/10.5334/dsj-2021-025 https://doi.org/10.3390/ai4040046 https://doi.org/10.5334/dsj-2024-047 https://github.com/streamlit/streamlit https://doi.org/10.1080/13658816.2019.1684500 https://doi.org/10.1145/3173574.3174045 https://doi.org/10.5334/cstp.227 https://doi.org/10.9781/ijimai.2024.02.003 https://doi.org/10.9781/ijimai.2024.02.003 https://www.mongodb.com https://www.mongodb.com https://www.authorea.com/doi/full/10.22541/essoar.171136837.71755629?commit=b0e4bff51da5dc2ec8f3e2e77ebc2a7427d04d1f https://www.authorea.com/doi/full/10.22541/essoar.171136837.71755629?commit=b0e4bff51da5dc2ec8f3e2e77ebc2a7427d04d1f https://doi.org/10.5334/dsj-2024-037 https://doi.org/10.5194/gi-11-195-2022 https://github.com/msu-denver/bili-core https://github.com/msu-denver/bili-core https://doi.org/10.3390/ai4040043 https://doi.org/10.2139/ssrn.4622514 https://doi.org/10.2139/ssrn.4622514 https://react.dev https://doi.org/10.48550/ARXIV.2408.04648 https://doi.org/10.1057/s41599-024-02720-3 https://doi.org/10.48550/arXiv.2408.14438 https://doi.org/10.48550/arXiv.2408.14438 https://arxiv.org/abs/2408.14438 [cs] Abstract 1 Introduction 2 Background and Motivation 3 Community Asset Mapping and Co-Design Process 4 Bili and BiliCore: A HCI-Informed Conversational AI Framework 5 Future Work 6 Conclusion Acknowledgments References