A data project in Bristol, England, shows how quickly public-sector analytics can move from coordination tool to source of public concern. The Think Family Database, launched in 2016 by Bristol City Council and Avon and Somerset Police, held records on close to half a million people who live in the city. For many years, few of them knew anything about it.
The database brought together sensitive information from several parts of public life. It included police intelligence reports, housing status, mental health records, teenage pregnancies, enrollment in parenting courses, and free school meals. Officials then built machine-learning models on top of that data to assign scores to thousands of adults and children.
What the Bristol system was built to do
The stated aim was to build what officials called a “picture of threat, harm, and risk” in the region. That goal sits at the center of the debate. Supporters saw the work as a way to help frontline services understand risk earlier and respond more efficiently. Critics saw a powerful system that could affect people without them knowing how they had been assessed.
At an event in early 2022 about tackling child exploitation crimes, one police data scientist described the approach in unusually plain terms: “I essentially dump all that data in a big bucket and stir it with a data-science spatula, and we come out with a lovely risk score for everybody.”
The Think Family Database was only one part of Avon and Somerset Police’s broader predictive analytics program. The force created at least 23 separate models. These included algorithms intended to identify the risk that people would commit burglary, fail to turn up in court, go missing, or become a victim of domestic abuse.
One senior officer described creating a “league table” of the area’s most dangerous criminals. That appears to refer to the Offender Management App, which was designed to hold data on around 300,000 people in the region.
Why residents struggled to understand the system
A central issue was visibility. The way Avon and Somerset Police developed and used predictive tools was not always clear to the public. John Pegram, who leads a local police accountability group in Bristol, said he did not hear about the Offender Management App until 2023, years after it had been created.
Pegram later suspected that he might be included in the system. “I think I knew I was on the app,” he says. In early 2024, he filed a request to learn how police were using his data. Police refused to say. Months later, after Pegram had hired solicitors, police confirmed he was on the app but declined to elaborate further.
That left open the most important practical questions. Pegram did not know whether an algorithm had scored him, what any score might be, or how the system could shape future interactions with authorities. The same uncertainty could apply to others in Bristol, the UK, and increasingly around the world where predictive analytics are being adopted by public agencies.
How the project grew from public-sector pressure
The roots of the program go back to a period of strain for Avon and Somerset Police. In 2014, the force was under pressure on multiple fronts. Like other UK forces, it had seen budgets cut. Its chief constable had been suspended. An official report had highlighted failures to follow procedures designed to protect some victims of domestic abuse.
After that report was published, the force’s head of performance said, “We believe predictive analytics is the solution.”
Gary Davies, a former police chief superintendent who had moved to Bristol City Council two years earlier, was also looking at data as a way to spot risk earlier. Davies led a council team supporting children and families. He said families already in crisis were easy to identify, but it was harder to see those near the top of a downward spiral.
His view was that different agencies each held pieces of the picture. A school might know about rising absences. Police might know that a child had recently witnessed domestic abuse for the first time. Separately, those details might not trigger intervention. Together, Davies argued, they could show a more serious trajectory.
Starting in 2015, a small group of Bristol City Council and Avon and Somerset Police staff worked together from one of the city’s police stations. The Insight Bristol team, headed by Davies, began combining data from across the public sector so frontline workers could see information about children and families in one place.
The consent and trust problem
The Insight Bristol team did not seek residents’ consent to use their data in the Think Family Database. Davies said the team relied instead on “legal gateways”, meaning data sharing considered necessary for agencies to meet legal obligations such as protecting children.
Initially, residents could not opt out of the database. Later, the council included an opt-out option in tax letters to residents. Davies, who recently retired, said the project did help protect children and made information about risk and vulnerability more efficient to use.
But the same facts that made the database useful to officials also made it sensitive for the public. The system gathered deeply personal records, made them available for risk assessment, and did so without many residents understanding the scale or mechanics of the project.
Davies said public communication was difficult because it was hard to get enthusiasm or interest from groups of people. He summarized feedback from those who did engage as: “We don't mind you using it to support us, but we don't want you to use it against us.”
What the documents showed
WIRED, working with Liberty Investigates, the Bristol Cable, and Lighthouse Reports, obtained hundreds of pages of documentation through public records requests. The reporting described the material as the most comprehensive picture to date of Avon and Somerset’s regional experiment with data collection and predictive analytics.
The investigation found that at least two risk-scoring models were quietly abandoned after Bristol City Council staff decided they could no longer trust them. Previously unreported documents also showed government inspectors and independent reviewers warning about a lack of transparency in parts of the program and the possibility that the systems could undermine public trust.
Police data disclosed to WIRED included more than 36,000 model performance scores. An independent analyst who reviewed the data for WIRED said that in some cases the scores appeared to show “genuinely poor predictive performance.”
The timing matters because the UK appears ready to use predictive analytics and artificial intelligence more widely across criminal justice. Andy Marsh, the former chief constable of Avon and Somerset, now heads the national standard-setting body for forces across England and Wales as CEO of the College of Policing. Marsh has said effective AI should be “injected like heroin” to speed up British police work.
In a recent interview, Marsh said the College of Policing was examining around 100 currently deployed AI tools, including tools for predictive policing. “Our job is to test the ones that work properly, test them with rigorous evaluation, and then spread them like wildfire through policing.”
The Bristol case shows why that evaluation matters. Predictive policing tools do not simply process data in the background. They can shape how public agencies see people, families, and risk. When the data are sensitive, the models are hard to inspect, and residents cannot easily learn how they are affected, the question is not only whether the technology works. It is whether the public can trust the system built around it.