Counting the Uncounted: How Community-Led Data Projects Are Redefining Police Violence Accountability in America

In 2014, the FBI's Supplementary Homicide Reports listed 461 people killed by law enforcement in the United States. In the same year, the Washington Post's independent count identified 990. The Guardian's concurrent database, The Counted, documented 1,146. The divergence was not a rounding error. It was a structural failure—one that human rights researchers recognized immediately as the kind of data problem they routinely encounter when documenting state violence in countries with weak accountability institutions.

A decade later, the federal government's voluntary reporting system remains the official record. And it remains, by nearly every independent assessment, a profound undercount.

The Federal Reporting Gap

The FBI's National Use-of-Force Data Collection program, launched in 2019 after years of advocacy pressure, was supposed to address the longstanding deficiencies in federal tracking. It has not. Participation is voluntary, and in its first full year of operation, fewer than 40 percent of law enforcement agencies submitted data. Agencies representing some of the largest urban police departments in the country—including those in New York and Los Angeles—initially declined to participate or submitted incomplete records.

The methodological problems run deeper than participation rates. The FBI collects data only on incidents that law enforcement agencies choose to self-report, using definitions and categories that agencies themselves apply. There is no independent verification, no cross-referencing with medical examiner records, and no systematic effort to capture incidents that departments classify under alternative categories. Civil rights attorneys and researchers have documented cases in which deaths during police encounters were recorded as accidents, medical emergencies, or suicides in official records.

This is precisely the kind of institutional data environment that international human rights organizations encounter when monitoring state violence in contexts with limited rule of law. The methodological responses they have developed over decades are now being adapted by American advocates.

Citizen Science Meets Human Rights Methodology

The most consequential police violence databases operating in the United States today were not built by government agencies. They were built by journalists, activists, data scientists, and bereaved family members who decided that the official record was insufficient and began constructing their own.

MappingPoliceViolence.org, developed by activist and researcher Samuel Sinyangwe, aggregates data from multiple sources—including local news reports, social media documentation, medical examiner records, and crowdsourced community submissions—to produce what its maintainers argue is the most comprehensive national database of police killings currently available. The project's methodology explicitly mirrors approaches used by international human rights monitors: triangulating across multiple independent sources, applying consistent definitional standards, and documenting its own limitations transparently.

Fatal Encounters, a database maintained by journalist D. Brian Burghart, has compiled records of over 35,000 deaths in law enforcement encounters since 2000—a dataset that predates and far exceeds what federal agencies have produced. Burghart's methodology involves systematic review of local news archives, obituaries, and public records requests across all 50 states. The project functions, in effect, as a distributed investigative newsroom dedicated to a single question: who has been killed by police, and under what circumstances?

The Washington Post's Fatal Force database and the Guardian's earlier Counted project demonstrated that newsroom-based monitoring, applying consistent definitional standards and independent verification, could produce substantially more accurate counts than voluntary federal reporting. Both projects documented racial disparities in lethal force that federal data had obscured or failed to capture.

Lessons from International Human Rights Monitoring

The methodological parallels between these domestic projects and established international human rights documentation practice are not coincidental. Several of the researchers involved in building US police violence databases have drawn explicitly on frameworks developed by organizations such as the Syrian Archive, the Human Rights Data Analysis Group (HRDAG), and the Patrick Ball–led statistical methods used in documenting atrocities in Guatemala, Kosovo, and Peru.

HRDAG's core contribution to human rights data science is the application of multiple systems estimation—a statistical technique originally developed in ecology to estimate total population size from incomplete samples. By comparing overlap between independent databases, researchers can generate statistically defensible estimates of the true total number of incidents, even when no single source is complete. HRDAG has applied this method to document killings by state security forces in Colombia, El Salvador, and Sierra Leone. Researchers affiliated with academic institutions have begun applying comparable techniques to US police violence data.

The Syrian Archive's approach to open-source verification—using geolocation, satellite imagery, and social media metadata to authenticate documentation of human rights violations—has direct domestic applications. In the United States, projects such as the Police Scorecard and the Invisible Institute's Citizens Police Data Project have incorporated open-source verification techniques to authenticate incident reports, cross-reference officer personnel records with complaint histories, and map patterns of excessive force at the precinct and individual officer level.

Community Documentation as Resistance

Beyond methodology, the community-based data projects operating across American cities reflect a political logic that is deeply familiar to human rights practitioners working in authoritarian contexts: when the state will not document its own abuses, affected communities must do so themselves.

In Chicago, the Invisible Institute has built one of the most detailed officer-level accountability databases in the country, incorporating decades of complaint records obtained through litigation against the Chicago Police Department. The project has identified officers with extensive sustained complaint histories who continued to receive promotions and faced no disciplinary consequences—a pattern of institutional impunity that the department's own records, not publicly available, had concealed.

In Minneapolis, community organizations began systematically photographing and video-documenting police encounters years before George Floyd's murder brought the department's practices to national attention. That documentation record proved critical to subsequent civil rights litigation and to the federal Department of Justice investigation that ultimately found a pattern and practice of constitutional violations.

In Los Angeles, the Youth Justice Coalition has trained community members to document stops, searches, and use-of-force incidents in real time, creating a distributed network of first-person accounts that supplements official incident reports. The methodology draws on witness documentation protocols developed by international human rights organizations for use in conflict zones—adapted for the specific legal and social context of American policing.

The Data Infrastructure Challenge

Despite their achievements, community-led data projects face persistent structural challenges. Funding is precarious and often dependent on foundation grants that follow issue-cycle attention rather than long-term institutional need. Staff turnover disrupts institutional knowledge. Legal challenges from law enforcement agencies and municipalities have targeted records requests and, in some cases, the databases themselves.

Interoperability remains a significant limitation. The multiplicity of independent databases—each applying somewhat different definitions, geographic boundaries, and data collection methods—makes aggregate national analysis difficult. Researchers at Stanford's Open Policing Project have worked to address this by standardizing data formats across jurisdictions, but the broader ecosystem remains fragmented.

Federal legislation that would mandate comprehensive, independently verified use-of-force reporting has repeatedly stalled in Congress. The George Floyd Justice in Policing Act, which included provisions for standardized national data collection, passed the House in 2021 and failed to advance in the Senate. Its data infrastructure provisions have not been revived in subsequent legislative sessions.

What the Data Already Shows

Despite these limitations, the existing community-built databases have produced findings of considerable evidentiary weight. Mapping Police Violence's analysis of data from 2013 through 2023 documents that Black Americans are killed by police at a rate approximately three times higher than white Americans—a disparity that persists after controlling for crime rates and other socioeconomic variables. The data further shows that the majority of police killings occur in situations that do not involve a violent crime, and that less than 2 percent of officers involved in killings face any criminal charge.

These are not advocacy talking points. They are empirical findings derived from systematic data collection, documented methodology, and independent verification. They meet the evidentiary standards that human rights organizations apply when documenting state violence anywhere in the world.

The community scientists, journalists, and advocates who built these databases did so because the official record failed them. In doing so, they have demonstrated something important: rigorous human rights data methodology is not the exclusive province of international institutions or well-funded NGOs. It can be practiced at the neighborhood level, with limited resources, by people who have a direct stake in the truth. That is, in many respects, exactly how the most consequential human rights documentation has always been done.