User:Wrigleyjillian/sandbox

Current techniques
Specialist software now exists for crime-prediction by analysing data.

This type of software has allowed law enforcement agencies to form predictions about criminal behavior and identify potential criminal hotspots based on crime data.

However, crime prediction software has also faced heavy criticism from academics, privacy, and civil liberties groups due to concerns about the lack of evidence proving the technology’s reliability and accuracy.

Another criticism of crime prediction software is that crime prediction algorithms often use racially skewed data in their analysis. This can potentially lead law enforcement agencies to make decisions and predictions that unfairly target and label minority communities as at risk for criminal activity.

A widely used criminal risk assessment tool called the Correctional Offender Management Profiling for Alternative Sanctions, or COMPAS, has been used by police and judges to predict the risk of recidivism amongst more than 1 million offenders since its development in 1998. The software predicts the likelihood that a convicted criminal will reoffend within two years based upon data including 137 of the individuals physical features and past criminal records.

A study published in Science Advances by two researchers found that groups of randomly chosen people could predict whether a past criminal would be convicted of a future crime with about 67 percent accuracy, a rate that was extremely similar to COMPAS.

Although COMPAS does not explicitly collect data regarding race, a study testing its accuracy on more than 7000 individuals arrested in Broward County, Florida showed substantial racial disparities in the software’s predictions.

The results of the study showed that black defendants who did not reoffend after their sentence were incorrectly predicted by COMPAS software to recidivate at a rate of 44.9%, as opposed to white defendants who were incorrectly predicted to reoffend at a rate of 23.5%. In addition, white defendants were incorrectly predicted to not be at risk of recidivism at a rate of 47.7%, as opposed to their black counterparts who were incorrectly predicted to not reoffend at a rate of 28%. The study concluded that the COMPAS software appeared to overpredict recidivism risk towards black individuals while underpredicting recidivism risk towards their white counterparts.