Overbooking’s Impact on Pre-Trial Risk Assessment Tools

A new paper, “The impact of overbooking on a pre-trial risk assessment tool,” published in the FAT* ’20: Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency,  addresses the question of how police officer booking decisions affect pre-trial risk assessment tools relied upon by judges to make pre-trial release decisions. The research is co-authored by Kristian Lum, San Francisco’s newly elected District Attorney, Chesa Boudin, and HRDAG executive director Megan Price, and it examines a popular risk assessment tool known as the Arnold Public Safety Assessment (PSA), using data from a pilot run of the tool in San Francisco, California.

The key finding of the paper is that booking charges that do not result in a conviction (that is, charges that are dropped or result in an acquittal) increased the recommended level of pre-trial supervision in around 27 percent of cases evaluated by the tool. In other words, it is fairly common that a person is booked on charges that are not sufficiently supported by the facts of the case to result in a conviction—or overbooked—and these charges alone cause the risk assessment tool to recommend to judges that the defendants have greater supervision before the trial.

“We were delighted to co-author this with Chesa, San Francisco’s newly elected DA, and we’re excited about continued collaborations with both the Public Defender’s Office and District Attorney,” says Megan.

Kristian presented the paper on January 29 at the Conference on Fairness, Accountability, and Transparency (FAT*) held in Barcelona.

FAT* is an annual conference dedicated to bringing together a diverse community to investigate and tackle issues such as machine learning, and the ethical, moral, social, and policy implications of big data and artificial intelligence.

Acknowledgments

Support for this project was provided by grants from the AI Initiative and the Ford Foundation. (For more information about HRDAG’s supporters, please see our Funding page.)

Image: This image, which was originally posted to Flickr, was uploaded to Commons using Flickr upload bot on  by Jacklee. On that date, it was confirmed to be licensed under the terms of the license indicated.