Data for Black Lives

"Data for Black Lives" is the movement name and governing message of a US-origin strand of the make-AI-good movement that frames data science as political terrain on which data scientists must choose sides — and argues that the same data systems deployed against Black communities can be reclaimed as tools of liberation, self-determination, and community power. The political argument lives in the preposition. "For," not "about" or "against": data science in service of, and governed by, the communities it affects rather than by the institutions that extract from them. In Yeshimabeit Milner's formulation, data is currently "a weapon of political oppression" operating through predictive policing, automated benefit denial, facial recognition, and algorithmic criminal sentencing; the organising premise is that it can instead be made a "tool for social change." Where the adjacent data is a civil rights issue framing routes the critique through civil rights law — accountability frameworks, institutional reform, rights claims against existing systems — "Data for Black Lives" names an affirmative programme: building community capacity to wield data, governing data resources from the community up, and training a generation of data scientists to orient their expertise toward liberation rather than toward the systems that extract from and surveil Black communities.

Origin

The message crystallised with the founding of Data for Black Lives (D4BL) at the MIT Media Lab in November 2017 by Yeshimabeit Milner and Lucas Mason-Brown in Cambridge, Massachusetts. Milner, a Brown University Africana Studies graduate who had organised communities in Miami and Chicago, had earlier used district-level data to document that Black children in Massachusetts school districts were being suspended at disproportionately higher rates than white children — an act of community-protective data analysis that the organisation's programme would later systematise. The founding premise, as Milner stated at the inaugural conference's opening panel, was that data and technologies "have far too often been weaponized against black communities" — and that the movement's task was to reverse the polarity, making data science responsive to community need rather than institutional control.

The inaugural conference (November 17–19, 2017) drew 400 participants, with 300+ on the waiting list watching by livestream, and established the distinctive convening model that would carry the message forward: data scientists, activists, and community organisers working together in the same room, identifying research agendas set by community need rather than institutional priority, and developing analytical tools for community use. The model was designed to dissolve the conventional boundary between the technical and the political — not to turn data scientists into advocates, but to insist that the choice of who a data scientist's work serves is already a political decision, made implicitly, in every methodological and employment choice a practitioner makes.

The core argument

The message rests on a claim about the positionality of data science. The standard data-ethics frame — including the machine bias critique — treats algorithmic discrimination as a technical failure to be corrected: a model is biased, an audit identifies the disparity, a technical fix is possible. The "Data for Black Lives" framing rejects the neutrality frame at the root. As Milner's "Abolish Big Data" manifesto argues, Big Data is "a philosophy; an ideological regime, one that determines how decisions are made and who makes them." Predictive policing algorithms, credit-scoring models, and automated welfare systems do not simply produce biased outputs that can be tuned — they are systems built to serve the interests of the institutions that deploy them, and the data scientists who build such systems are making choices about whose lives their technical labor serves.

The affirmative inversion follows directly: data science expertise — the same technical capacity that powers systems applied against Black communities — can be placed at community disposal. D4BL's organisational mission names both sides of the opposition: "a movement of activists, organizers, and scientists committed to the mission of using data to create concrete and measurable change in the lives of Black people." The governance corollary is the redistributive demand: communities that are currently subjects of data collection and algorithmic decision-making should be able to set research agendas, control the terms of data collection about themselves, and have access to the analytical tools needed to understand and contest the systems that govern their lives. Milner's "Abolish Big Data" formulates this demand as placing "data in the hands of people who need it the most" — community data sovereignty as the substantive goal, not technical remediation of disparate impact.

This is the distinctive register of "Data for Black Lives" as a message. The data is a civil rights issue framing insists on accountability — existing data systems must be subjected to civil rights law, community members have rights claims against the systems that harm them. "Data for Black Lives" insists on agency — community members should govern the data systems that affect them, and data scientists should orient their technical work toward community benefit. The two framings are complementary rather than contradictory, but they name different action programmes: accountability routes toward litigation, regulation, and institutional reform; the "for" framing routes toward community data capacity-building, participatory research design, and the political education of data scientists.

Propagation

D4BL propagated the message through its annual conference series at MIT, growing to a network of over 20,000 scientists and activists across the United States. The organisation launched regional chapters — beginning with Pittsburgh — and developed programme work spanning Abolition, Political Education, Data Weapons, Data Governance, Algorithms, and Democracy, demonstrating the participatory research model across distinct issue domains. The #No More Data Weapons campaign (launched February 2021) carried the weapon-versus-tool opposition into a targeted campaign against specific surveillance technologies — predictive policing tools, facial recognition, social-media monitoring — naming the dual-use character of data systems directly in its campaign framing and demanding that institutions stop "investing in data weapons, building new data weapons, [and] disguising data weapons as legitimate and neutral."

The participatory model spread into the wider field through organisations that developed within D4BL's intellectual orbit. The Distributed AI Research Institute (DAIR), founded by Timnit Gebru in Oakland in December 2021, operationalised the community-rooted research model as an independent institution: "grounded in community expertise and local needs" and explicitly set against the agenda-setting authority of institutional funders. DAIR inherited D4BL's insistence that research agenda-setting be responsive to affected communities — the "community-rooted" frame is a direct heir to the "data for" orientation. The Algorithmic Justice League connects the participatory governance emphasis to the specific technical domain of facial recognition auditing, building community feedback on algorithmic harm into the audit methodology and working within the same scientist-activist bridge that D4BL established.

The Movement Scientist Fellows Program, which D4BL launched in 2024, is the most direct institutionalisation of the message's organising core: training activists and organisers in data analysis for social change, reversing the direction of data expertise from institution-to-community to community-controlled. The MacArthur Foundation's operational grant to D4BL (2019–2021) confirmed the framing as a legible philanthropic category distinct from the civil rights accountability register and from the technical algorithmic-fairness research register — two existing institutional homes for adjacent work; the "data for" community capacity-building register required a new kind of organisation the MacArthur grant recognised as fundable.

The indigenous data sovereignty movement developed in parallel across Anglophone indigenous communities in the same period, articulating a related but distinct claim: that indigenous communities have inherent rights to govern data about themselves, their lands, and their peoples, grounded in sovereignty rather than in the US civil rights tradition. The two framings converge on community data governance as the substantive demand while diverging on the political and legal traditions they draw on; the convergence makes "Data for Black Lives" part of a wider global articulation of community data sovereignty that the movement is still actively building.

Why it has carried

Three features explain the message's traction.

First, the "for" framing holds an affirmative programme that accountability critique alone does not. The machine bias framing names a harm and demands correction; the data is a civil rights issue framing routes the demand through civil rights law and institutional reform. Both are fundamentally reactive — they answer "what is wrong with existing systems?" "Data for Black Lives" answers instead "what should data science be for?" — and that affirmative answer organises an action programme (build community capacity, set community-governed research agendas, train data scientists oriented toward liberation) that does not depend on the reform of existing institutional systems. The distinction matters organisationally: accountability work requires engaging institutional gatekeepers who can refuse; community capacity-building can proceed without those gatekeepers' permission.

Second, the scientist-activist convening model gave the message organisational machinery with a distinctive political economy. Data scientists are typically inside institutional structures — companies, universities, government agencies — that determine what work gets funded and deployed. D4BL created an institutional space outside those structures where data scientists could encounter the communities their work affects and make different choices about whose interests they serve. The inaugural 400-person conference built from a Twitter account in eight months, with 300 more on the waiting list, demonstrated that the demand for that space was latent and large. The DAIR Institute's founding six years later with the same community-rooted orientation and immediate institutional funding from the Ford Foundation, MacArthur Foundation, Kapor Center, and Open Society Foundations confirmed that the model was replicable and fundable outside D4BL's specific organisational context.

Third, the message travels across the data governance landscape as a frame for community agency that does not require adopting any particular policy or technical position. "Data for" is a governance orientation — who should the data system serve, and who should set the research agenda — that applies to predictive policing, welfare algorithms, health data governance, credit systems, and AI safety without requiring the same substantive analysis in each domain. D4BL's programme breadth — Abolition, Political Education, Data Weapons, Data Governance, Algorithms, Democracy — is the organisational evidence that the frame is domain-general: it spans any arena in which data systems govern Black lives, and by extension any arena in which communities are currently data subjects rather than data agents.