Algorithms are increasingly coming to trial in the United States, where they help determine the risk of re-offending … and decide whether to be jailed or released. Predictions all too often skewed by irrelevant historical data, reports MIT Technology Review magazine .
Faced with the imperative of reducing the number of prisoners without increasing the crime rate, the American judicial system is turning more and more to technological tools to try to improve its efficiency.
After the crime-predicting algorithms for the police, it is now that artificial intelligence claims to be able to give the accused a score in court. This score, obtained by collecting data on the person’s profile and historical data, would reveal the risk that the accused would pose if released.
It is then up to the judge to take into account or not this score in his decisions: choice of the services of recovery, imprisonment or not during the trial, severity of the sentence, etc.
People favoring the use of artificial intelligence by the courts believe that algorithms could reduce or even eliminate the partiality of judges by basing decisions solely on data.
Biased algorithms denounce opponents
But it is precisely the question of partiality that upsets the opponents of this technology. Rather, they believe that these algorithms are inherently biased, since they rely on historical data to make certain predictions.
Tools like this try to find correlations in large amounts of data. As low-income communities and minorities have historically been more targeted by law enforcement, artificial intelligence tools are more likely to conclude that a person from these communities poses a risk to society, explains the MIT Technology Review.
However, the correlation between two categories of data does not mean that there is a causal link. By basing their conclusions on biased data, the algorithms would therefore tend to magnify these biases, fueling a vicious circle that would penalize the judiciary communities.
Despite the call, in July (New Window), the American Civil Liberties Union and the National Association for the Advancement of Colored People, American organizations for the defense of civil liberties, to turn their backs on this technology, moreover more states are beginning to use it, hoping to reduce the occupancy rate of their prisons.