An interactive demonstration of how the justice system uses data to determine the length of prison sentences
When deciding how long to send someone to jail, many states currently use statistical models to determine whether offenders risk committing a future crime if they are let out on probation or parole. In the past several years, researchers have been able to demonstrate that factors like drug and alcohol problems, family life and education can help them predict the likelihood of recidivism.
In a speech before the National Association of Criminal Defense Lawyers Friday, Attorney General Eric Holder warned that this increasingly popular use of data-based methods in determining prison sentences “may run the risk of imposing drastically different punishments for the same crimes.” As Holder told TIME this week, he fears that the statistical methods that punish for factors like education will disproportionately affect minority and poor offenders.
Below, you’ll find a demonstration of the kind of kind calculator many states use to predict odds of recidivism. Change the responses in the following interactive to see how the odds of re-arrest change with the offender’s circumstances. In many states, these odds are being used to determine sentencing lengths.
The actual use of this “post conviction risk assessment” varies widely. This method, developed by criminal justice researcher Christopher T. Lowenkamp and colleagues, is an area of ongoing study. Using standard statistical models, the researchers were able to study a large population of offenders to determine which factors can predict a person’s likelihood of future offense and which cannot. Notably, a person’s race–left in this interactive for demonstration purposes–has almost no predictive power over future behavior when all other factors are held constant. In other words, a white offender and black offender with the same answers to the above questions are almost equally likely to commit a future crime.