Sign In

Communications of the ACM

ACM TechNews

Partnership on AI: Algorithms Aren't Ready to Automate Pretrial Bail Hearings


View as: Print Mobile App Share: Send by email Share on reddit Share on StumbleUpon Share on Hacker News Share on Tweeter Share on Facebook
A human judge oversees a bail hearing.

The first-ever report from the Partnership on AI (PAI) declares the algorithms now in use to automate the pre-trial bail process unfit to do so.

Credit: Kin Man Hui

The Partnership on AI (PAI) has released a report declaring the algorithms now in use to automate the pre-trial bail process unfit to do so, as they suffer from data sampling bias and bias in statistical predictions, as well as human-computer interface issues and unclear definitions of high risk and low risk.

In its first-ever report, the organization recommends policymakers avoid using algorithms entirely for decision-making surrounding incarceration, or find ways to meet minimum data collection and transparency standards.

From Venture Beat
View Full Article

 

Abstracts Copyright © 2019 SmithBucklin, Washington, DC, USA


 

No entries found