Predictive police: AI risks increasing discrimination by copying old crime cards

The Adrian Dredge trial, commissioned by the Association of British Science Writers, shows how biased police data feed into a loop where the algorithm confirms where we've already looked.

The young British science author Adrian Dredge has won the 2026 Young Science Writer Award from the Association of British Science Writers with an essay, published by the BBC, on the possible excesses of AI in policing. His central image: a cartographer locked up without a window, who is only entrusted with fifty-year-old plans. Tasked with predicting where the next crimes will take place based on past crimes, he shades the same neighborhoods; the police patrol more, therefore see more offenses, and the next map darkens even more. The statistics then reflect both surveillance choices and real crime.

"Dirty" data

The essay is based on the work of sociologist Sarah Brayne, who has observed the use of big data by the Los Angeles Police Department from the inside and describes seemingly neutral predictions that help produce the events they anticipate. Researchers at New York University also found that in nine of the thirteen U.S. jurisdictions studied, data from periods of illegal or biased practices could be used to train these systems. Carolyn McKay, from the University of Sydney, argues for independent human control of data before any use. Dimitrios Kollias, from Queen Mary University of London, nuances: modern models do not just copy their examples, but it is all the more important to ensure that they remain under real human control.