Experts have developed an artificial intelligence that can predict any arrest within three years of a prisoner’s release on parole. The machine learning is designed to identify the risk of releasing a prisoner early by analyzing 91 variables, including age, race, and previous arrests.
Scientists from the University of California used data from more than 19,000 inmates scheduled with the New York State Board of Parole from 2012 to 2015.
Court documents show 4,168 people were released, but the AI determined that the council could have released twice as many prisoners without increasing the subsequent arrest rate.
A machine learning algorithm found that the expected risks for those who had been denied parole and those who had been released were very similar.
“We conservatively estimate that the board could have doubled the release rate without increasing the overall arrest or violent felony rate, and they could have achieved these gains while simultaneously eliminating the disparities Ethnic release rates.
There were a total of 19,713 people on parole hearings from 2012 through 2015. During these years, 4,561 individuals were paroled, and were rejected 16,068 times or more, but combined with factors such as age and race, The AI also looked at specific crimes, such as those related to minors, drugs, hate crimes, and theft, and was given information regarding arrest history, and predicted crime rates on parole.
And while human parole staff have used statistical analysis for decades to determine whether an inmate should be released, this technology is able to quickly look at every variable that can be “knowable” for certain.