Sign In

Communications of the ACM

ACM TechNews

Are Programs Better Than People at Predicting Reoffending?


A prisoner on trial.

The COMPAS computer program performs about as well as humans at predicting recidivism, according to a Dartmouth College research study.

Credit: Getty Images

A Dartmouth College research study found the COMPAS computer program performed about as well as humans at predicting recidivism.

The researchers had 400 volunteers at Amazon Mechanical Turk, and COMPAS, predict who out of 1,000 defendants chosen at random would be arrested for another crime within two years of arraignment. Each volunteer saw only one group of 50 subjects, and each group was seen by 20 volunteers.

The volunteers correctly predicted whether someone had been rearrested 62.1% of the time, which climbed to 67% when the assessments of the 20 who studied a particular defendant's case were pooled. Meanwhile, COMPAS scored a recidivism rate of 65.2%, and a follow-up experiment in which the defendant's race was mentioned made no difference.

The implication is that COMPAS is as competent as human common sense at parsing relevant facts to predict who will and will not be rearrested.

From The Economist
View Full Article - May Require Free Registration

 

Abstracts Copyright © 2018 Information Inc., Bethesda, Maryland, USA


 

No entries found