This algorithm predicts crimes a week in advance!

American researchers have created an Artificial Intelligence capable of predicting the crime rate a week in advance in several cities in the United States, with an accuracy of 90%. Far from wanting to identify criminals in advance, they rather hope to identify police bias.

The dystopian world of film Minority Report, where criminals are caught before committing a crime, may well become a reality one day. However, it will not be with gifts like precognition, but rather relying on Artificial Intelligence. In an article published in the journal Nature Human Behaviorresearchers at the University of Chicago have developed an AI capable of predicting the level of crime a week in advance within a radius of 300 meters, with an accuracy rate of 90%.

The researchers analyzed crime data from 2014 to 2016 in Chicago by dividing the city into a grid where each square is about 300 meters on a side. They then successfully predicted crime levels for each square for the following weeks. They obtained similar results after repeating the operation with seven other major cities in United States.

An AI susceptible to racial bias?

This kind of algorithm has been criticized in the past because biases racial. The chicago police had tested an algorithm to identify the people most at risk of being involved in fire, as aggressor or victim. The problem? 56% of black men aged 20-29 were on the list…

The researchers indicate that, in their case, we are far from Minority Report because their algorithm only predicts crime locations, not the people involved. Moreover, according to them, their tool can be used to analyze police bias. In particular, they noted that crimes committed in affluent neighborhoods resulted in arrests more often than in poor neighborhoods, highlighting a biased police response.