Community-driven data science practices

Mathematics researchers are becoming more involved with research questions at the interface of data science and social justice. This type of research needs to be grounded in the needs of the community in order to have significant impact. In this paper, we examine two examples of community-research partnerships in data science for social justice co-authored by both community members and mathematical researchers. The first, VECINA, is a place-based community-research partnership focused on environmental justice. VECINA introduces a framework for developing fruitful local collaborations. The second example, SToPA, originates in citizens’ request for an analysis of their town’s policing data, but focuses on how to scale this work beyond that place-based setting. SToPA’s research helps us imagine how we can continue to actively collaborate with community members even when working to scale projects beyond a single community. In both of these case studies, we examine the harmonies between established principles of power, process, and perspective with our framework for research-community partnerships. We use a duoethnography approach, directly illustrating the experiences of researchers. We also offer a set of reflections on the impact of these research-community partnerships.

💡 Research Summary

This paper presents a concrete, practice‑oriented framework for community‑driven data science research, illustrated through two case studies: VECINA (Visualizing Environmental and Community Information for Neighborhood Advocacy) and SToPA (Small Town Policing Accountability Lab). Both projects were co‑authored by mathematicians, data scientists, statisticians, and community activists, and they aim to embed social‑justice concerns—environmental justice, policing accountability, climate resilience—into quantitative research.

The authors foreground three guiding principles—Power, Process, and Perspective—derived from prior community‑based research literature (Bonner, Anti‑Racist Community Engagement, Design Justice, etc.). “Power” stresses that research questions, data selection, and outcomes must be shaped by the community’s priorities and capacity to effect change. “Process” calls for long‑term, relational engagement rather than transactional, short‑term collaborations. “Perspective” insists that lived experience and disciplinary expertise be treated as equally valid sources of knowledge.

Figure 1 (referenced in the text) visualizes how these principles map onto a step‑by‑step partnership workflow: (1) co‑identifying problems, (2) jointly designing data collection and analysis, (3) iterative feedback loops, (4) co‑producing visualizations or statistical reports, and (5) translating findings into policy or community action. The framework is positioned as an adaptation of the Bonner Community‑Based Research model, enriched with anti‑racist and design‑justice lenses.

In the VECINA case, the partnership began at ICERM (Brown University) and continued at the Institute for Mathematical and Statistical Innovation (IMSI, University of Chicago). Researchers collaborated with the Woonasquatuck River Watershed Council and the “Nuevas Voces” program in Olneyville, Providence, a historically marginalized, predominantly immigrant neighborhood. Through workshops, climate‑resilience training, and participatory mapping, residents helped define flood‑risk metrics, greenhouse‑gas emission sources, and green‑space needs. The duo‑ethnography method captured both researcher reflections and community narratives, revealing how power asymmetries were negotiated and how the “objective” veneer of data science was reframed as a tool for advocacy.

The SToPA project originated from citizen demands for transparency in policing data in Williamstown, MA, and later expanded to Durham, NC. The lab assembled data scientists, activists, and statisticians to clean, standardize, and analyze stop‑and‑search, use‑of‑force, and complaint records. Recognizing the challenges of scaling a place‑based effort, the team formed the Data Science, Police Accountability, and Community Engagement (DSPACE) group to develop reusable pipelines, ethical review protocols, and community‑led interpretation guides. The case demonstrates how the same three principles can guide both localized and broader‑scale interventions, emphasizing the need for adaptable governance structures when moving beyond a single community.

Both case studies discuss concrete obstacles: limited funding, restricted data access, institutional inertia, and the risk of reproducing epistemic injustice when academic voices dominate. The authors argue that transparent communication, shared decision‑making, and sustained capacity‑building (e.g., training residents in data literacy) are essential mitigations. They also note that publishing academic papers is only one output; real impact is measured by policy changes, community empowerment, and the creation of enduring networks.

The paper concludes with a set of reflections: (1) authentic co‑creation leads to research that is more likely to be used by communities; (2) duo‑ethnography is a valuable reflexive tool for surfacing power dynamics; (3) scaling requires balancing local specificity with standardized, ethically vetted data practices; (4) long‑term partnerships demand ongoing educational components and clear pathways for translating findings into actionable change.

Overall, the manuscript offers a rigorously documented, theoretically grounded, and practically tested roadmap for mathematicians and data scientists who wish to align their technical expertise with social‑justice goals through equitable, community‑centered research partnerships.