Algorithms that change lives should be trialled like new drugs

An algorithm used by US courts to predict reoffenders turns out to be no more accurate than random people on the internet. Why wasn’t it properly tested before now?