Civic AI Applications Lab
The Civic AI Applications Lab brings together a diverse range of research projects that extend beyond the more focused agendas of the Center's other labs. Rather than concentrating on a single topic or methodology, it provides a home for interdisciplinary research that applies AI to challenges in different domains or examines its broader implications for society.
The lab supports faculty-led projects grounded in their respective disciplines while creating opportunities for collaboration across fields. As related projects and collaborations develop, they may evolve into more focused research clusters and eventually form dedicated labs within the Center.
Research Projects
Public Health
Understanding Medication Diversion in Jails Using AI
This project uses natural language processing (NLP) and large language models to study medication diversion in correctional settings, with a focus on treatment for opioid use disorder. Researchers from Tufts University School of Medicine, NYU Grossman School of Medicine, and Tufts Computer Science are collaborating with the Middlesex County House of Corrections to analyze incident reports, identify patterns of medication diversion, including mutual exchanges and coercive circumstances, and examine institutional responses. The research combines AI-assisted classification with human qualitative analysis to better understand the circumstances surrounding diversion and inform evidence-based approaches to correctional healthcare.
Contact: Thomas Stopka
Using AI to Analyze Qualitative Data on the Opioid Crisis
This project investigates how large language models (LLMs) can support qualitative research on the opioid crisis, particularly among people who use drugs with interactions with the criminal-legal system. Drawing on interviews from the Justice Community Opioid Innovation Network study in Massachusetts, the research examines experiences with substance use and overdose, barriers to treatment, and attitudes toward medications for opioid use disorder among incarcerated individuals. The project evaluates whether LLMs can reliably identify themes and patterns in these interviews by comparing AI-generated analyses with those conducted by human researchers. By examining different models and approaches to domain-specific training, the study aims to develop methods for more timely and reliable analysis of qualitative evidence, helping public health researchers better understand evolving challenges and inform responses to the opioid crisis.
Contact: Jack Cordes
Mental Health and Overdose Prevention Explorer (mHOPE)
The Mental Health and Overdose Prevention Explorer (mHOPE) is an interactive, web-based platform designed to improve access to data on opioid-related overdoses, mental illness, treatment availability, prevention efforts, and social determinants of health. Through interactive GIS maps, trend visualizations, and resource mapping, the dashboard helps researchers, healthcare providers, policymakers, community organizations, family members, and peers to identify geographic disparities, understand evolving public health challenges, make informed decisions about allocating resources, and identify treatment options. The project also aims to provide educational resources for healthcare professionals and community leaders in shaping the platform. Future work will explore integration of AI to support natural-language queries, identification of patterns across complex datasets, and make public health information more accessible and actionable.
Contact: Thomas Stopka
Arts & Culture
Ways of Knowing
Ways of Knowing is an artistic research project examining how generative AI shapes our understanding and representation of the world. Between 1892 and 1963, the Keystone View Company sold 3D photographs of remote landscapes — glaciers, canyons, mountain vistas — each paired with a handwritten description, promising viewers the experience of a place they'd likely never visit. This project strips away the photographs and feeds only the words into generative AI video tools, asking machines that never saw the place to render someone else's century-old description as photorealistic footage. The compression happens twice: a century of change collapsed into one imagined scene, and the underlying computation itself compressed, run in parallel across thousands of processors in seconds. Keystone promised you could hold a distant place in your hand; AI promises a place conjured whole from a paragraph. Both are fantasies of total capture, and both eventually fail — but where Keystone's limitations (its colonial framing, its selective gaze) are easy to spot in hindsight, AI's are hidden by design. The project asks what happens when one exposed fantasy of total capture is used to try to resurrect another whose failures we can't yet see.
Contact: Mat Rappaport
Project Website: Ways of Knowing
AI Governance & Accountability
Holding AI Firms to Account: Roles of Governing Boards
Amid the highly politicized discourse on AI regulation, much less attention has been devoted to the corporate governance of AI firms themselves and particularly to the roles and duties of boards of directors. How can “governance” and “accountability” be conceptualized for the internal oversight of AI firms — and is there anything conceptually distinct about them that may require new approaches to governance? This project begins with the premise that organizational governance is foundational to public accountability — for it is the role of governing boards to address the three core questions of accountability that confront any organization and its role in society: for what is the organization accountable, to whom does it owe accountability, and how is it to exercise that accountability? This project aims to unpack these questions for AI firms, and to examine whether public accountability is (or should be) conceptually distinct in such firms as compared to nonprofit charities, for-profit businesses, or hybrid social enterprises. The second part of the research examines the governance structures of AI firms, with special attention to: legal status, ownership, voting power, and board composition and independence. Through empirical work, this project seeks to understand how governing boards grapple with the tensions between safety and innovation, and the terrain between private benefit and public purpose.
Contact: Alnoor Ebrahim