London Metro Implements Bias Recognition System for Black Commuters

At a meeting of the Parliamentary Committee on Science, Innovation and Technologies of Great Britain, it was stated that the person of recognition of persons used by the London metro during the coronation of the king may show racial bias with certain threshold values. On May 6, 2023, the coronation of the British king Charles III took place in London.

Dr. Tony Mansfield, the chief researcher of the National Physical Laboratory, told the deputies that the system based on the product of the company NEC and used by the country’s largest police department is prone to bias against black people created for its research.

It was found that if the system works on low and light thresholds of coincidence, then it begins to show the bias against black men and women together. Mansfield believes that the metro does not use the system at such threshold values.

According to a doctor’s report , false positive identities increases with low thresholds persons (0.58 and 0.56) and “begin to show a statistically significant imbalance between demographic groups: blacks have more false positive works than Asian or white.”

However, the NEC Neoface system with higher thresholds (0.64 or higher) did not give false positive identifications and, therefore, had no bias. With an average threshold (0.60 – 0.62), the system gave 8 false positive results, but did not have a statistically significant bias on demographic signs.

The Metro representative said that after the publication of the study, the system was used 3 times, including in some areas of London and during the coronation of the king.

During the operations of the police, in which 6 police van were used, there were 4 alerts, and there were no false positive results at all. There were 2 arrests, and according to the remaining notifications, the police decided that the arrest was not required, although the identification was correct.

At each operation, the police created a separate set of data for the “observation list” based on the location and “reconnaissance” for the operation. After each operation, the observation list was removed. No data from persons to recognize persons in real time were stored, according to the representative of the subway.

/Reports, release notes, official announcements.