In an era where algorithmic decision-making increasingly determines who receives social services, housing assistance, and other critical support, Dr. Francisca García-Cobián Richter and collaborators are working to ensure that the human stories behind the data aren’t lost in translation. Her innovative FAIR2 framework represents a fundamental shift in how researchers approach data analytics, one that prioritizes community knowledge alongside statistical rigor.
Richter’s work at the Mandel School of Social Work emerged from a growing recognition that traditional data science, despite its statistical sophistication, often misses crucial context about the people it purports to serve. “There’s nothing that can supply the qualitative description of how that data is generated,” she explains. “What are the experiences of people that are represented in the data?”
This insight led to the development of FAIR2, an extension of the widely-used FAIR framework for data management. While FAIR focuses on making data findable, interoperable, accessible, and reproducible, Richter’s addition emphasizes the need to frame data with community input, identify potential biases, and report findings back to affected communities.
The framework addresses a persistent problem in social policy research: the disconnect between researchers who analyze data and the communities whose lives generate that data. Traditional approaches often treat demographic information as fixed categories, failing to account for the complex realities of how people experience and report their identities over time.
“When we see that maybe a person represented in our longitudinal data might be reported with various types of races or even various types of genders over time or across data systems, what does that mean?” Richter asks. Rather than making assumptions or forcing people into predetermined categories, FAIR2 encourages researchers to engage directly with communities to understand these variations.
This approach becomes particularly critical when examining integrated data systems, which combine information from multiple government agencies. Richter’s center manages one of the country’s most comprehensive county-wide integrated data systems, linking 35 different agency databases. While these systems enable powerful analyses of social programs and outcomes, they also risk perpetuating existing biases if not carefully managed.
The FAIR2 framework addresses these limitations through what Richter calls “data chats”, structured conversations with community members who can provide context about their experiences with data collection and social services. These sessions aren’t just about gathering opinions; they’re about understanding the material conditions that shape how data is created in the first place.
One compelling application involves people experiencing homelessness who interact with coordinated intake systems. Rather than simply analyzing their demographic data, FAIR2 encourages researchers to understand how the experience of providing that information shapes both the data quality and the individuals’ interactions with services.
This community-centered approach extends to Richter’s teaching philosophy. The FAIR2 framework was originally developed as an educational tool for a new Data Science for Social Impact Certificate program. The goal isn’t to turn social workers into statisticians, but to equip them with tools to meaningfully engage with algorithmic systems that increasingly influence their practice.
“We don’t need social workers to be experts in data science,” Richter explains. “With the FAIR2 framework, they are able to analyze and bring in the voices of people that are represented in the data, understand the general assumptions that are used and question them.”
This approach has particular relevance as AI and machine learning systems become more prevalent in social services. Algorithms increasingly determine who gets priority access to homeless services, which child welfare calls are screened for investigation, and how social programs allocate resources. Without community input, these systems risk codifying existing inequalities under the guise of objective analysis.
Richter’s work with engineers reveals both the potential and challenges of bridging technical and social perspectives. Many engineers express reluctance to engage with ethics, claiming they lack expertise in moral philosophy. FAIR2 sidesteps this resistance by focusing on data quality, improved metadata and assumptions, rather than abstract ethical principles.
“There’s a data generating process underneath all the social data that we collect, we should learn more about,” Richter explains. “Rather than exclude that knowledge, exclude those individuals that have collective expertise, we make an effort to include that and include those insights.”
This framing resonates with technical practitioners because it speaks to professional concerns about data quality and analytical rigor. Just as engineers wouldn’t use materials without understanding their properties, FAIR2 argues they shouldn’t use social data without understanding how it was created and what it represents.
However, adoption remains challenging. Richter acknowledges that many engineers working on public interest technology feel isolated within their departments, lacking institutional support for incorporating community perspectives into their work. The framework’s success will likely depend on broader cultural shifts within both academic institutions and technology companies.
The stakes of this work extend far beyond academic methodology. As algorithmic systems become more sophisticated and pervasive, the need for community-informed data science becomes increasingly urgent. FAIR2 offers a practical roadmap for ensuring that technological advancement serves rather than supplants human wisdom about social challenges.
FAIR2 embodies a key insight: that truly effective social policy requires not just better algorithms, but better relationships between researchers, technologists, and the communities they serve. In an age of big data and AI, FAIR2 reminds us that the most important insights often come from simply asking people about their own experiences and actually listening to the answers.
By institutionalizing community knowledge within data science workflows, FAIR2 represents more than a methodological improvement; it’s a step toward more democratic and accountable forms of technological governance. As cities and social service agencies increasingly rely on algorithmic decision-making, frameworks like FAIR2 may prove essential for ensuring these systems truly serve the public interest.
— Mythili Sampathkumar