[Daily Mail, where America gets its news] - Scientists trained the AI to analyze 91 variables to determine if an inmate is at high or low risk of rearrests after being released on parole
...which means the AI was given the programmers’ prejudices and assumptions.
Scientists from The University of California, Davis
- The variables include age, race, sentences and previous arrests
- The AI was tested on data from the New York State Parole Board from 2012 to 2015, a time when more than 19,000 inmates had parole hearings
- More than 4,000 were released, but the AI found the parole board should have released nearly 50 percent more
Can we assume the programmers hail from the left side of the spectrum?
|