The **Think Family Database**, launched in 2016 by the **Bristol City Council** and the regional **Avon and Somerset Police**, has quietly amassed a vast repository of sensitive personal data on close to half a million Bristol residents. This includes police intelligence reports, housing status, mental health records, and even records of teenage pregnancies and enrollment in parenting courses. ### Algorithmic Risk Scoring and Opaque Operations Beyond mere data collection, officials employed machine-learning models to assign risk scores to thousands of adults and children. The stated goal was to build a “picture of threat, harm, and risk” within the region. As one police data scientist put it, the process involved essentially dumping all data into a “big bucket” to “stir it with a data-science spatula” and derive a “lovely risk score for everybody.” This risk scoring was part of a broader predictive analytics program by **Avon and Somerset Police**, which developed at least 23 separate models. These algorithms aimed to predict various outcomes, from the likelihood of committing burglary or failing to appear in court, to becoming a victim of domestic abuse. One senior officer even spoke of creating a “league table” of dangerous criminals, referencing the **Offender Management App**, designed to hold data on approximately 300,000 individuals.  ### Public Ignorance and Transparency Failures The public remained largely unaware of these sophisticated tools. **John Pegram**, a leader of a local police accountability group in Bristol, only learned about the **Offender Management App** in 2023, years after its creation. His attempts to ascertain how his own data was being used were initially met with refusal by the police, who only confirmed his inclusion on the app months later after legal intervention. In collaboration with **Liberty Investigates**, the **Bristol Cable**, and **Lighthouse Reports**, WIRED conducted an extensive investigation, obtaining hundreds of pages of documentation through public records requests. This has provided the most comprehensive insight to date into **Avon and Somerset**’s predictive analytics experiment. ### Abandoned Models and 'Poor Predictive Performance' The investigation revealed that at least two risk-scoring models were quietly abandoned after **Bristol City Council** staff deemed them untrustworthy. Previously unreported documents also show government inspectors and independent reviewers flagging a significant lack of transparency and warning that these systems could erode public trust. An independent analyst, reviewing over 36,000 model performance scores disclosed to WIRED, found evidence of “genuinely poor predictive performance” in some cases. ### A National Push for AI in Policing These revelations come as the UK appears increasingly ready to adopt predictive analytics and artificial intelligence across its criminal justice system. **Andy Marsh**, former chief constable of **Avon and Somerset** and now CEO of the **College of Policing**, is a vocal proponent. Marsh advocates for injecting effective AI “like heroin” to accelerate police work, stating his organization is evaluating around 100 deployed AI tools, including for predictive policing. His goal is to “test the ones that work properly, test them with rigorous evaluation, and then spread them like wildfire through policing.” ### The Genesis of Predictive Policing in Bristol **Avon and Somerset Police** faced significant pressures in 2014, including budget cuts and criticism over its handling of domestic abuse cases. **Gary Davies**, then head of performance, declared predictive analytics to be the “solution.” Around the same time, Davies, who had moved to the **Bristol City Council**, saw data as a way to identify families on a “downward spiral” before crises escalated. In 2015, the **Insight Bristol** team, led by Davies, began consolidating data from various public sector sources to create a holistic view of children and families. This led to the creation of the **Think Family Database**. ### Consent and Public Perception The **Insight Bristol** team did not seek direct consent from residents for data inclusion. Instead, they relied on “legal gateways,” asserting that data sharing was necessary to fulfill legal obligations, such as child protection. Initially, no opt-out was available, though one was later introduced via tax letters. Davies, now retired, believes the project successfully improved the understanding of risk and vulnerability for children and families. He noted that public engagement on the matter was difficult, but those who did engage generally understood the need for data use, with feedback summarized as: “We don't mind you using it to support us, but we don't want you to use it against us.”