acm-header
Sign In

Communications of the ACM

ACM TechNews

U.S. Prisons Mull AI to Analyze Inmate Phone Calls


View as: Print Mobile App Share:
hand holding cellphone behind prison bars

The U.S. House Appropriations Committee's push to study the use of artificial intelligence (AI) to analyze prison inmates' phone calls has prisoner advocates and families warning of risks of error, misunderstandings, and racial bias. Several state and local prisons have already begun using such technology, which can automatically transcribe inmates' phone calls and flag certain words or phrases. The House panel is urging the Department of Justice to consider potential federal utilization and to identify shortcomings in the information the tech generates.

The Oxford, AL police department has deployed Verus software from LEO Technologies, which uses Amazon Web Services’ natural language processing and transcription technology to process and flag prisoner calls. Oxford Police Chief Bill Partridge says the surveillance has helped local forces solve cold case homicides and has prevented suicides.

Critics warn of tools potentially amplifying racial bias; for example, a Stanford/Georgetown University analysis found Amazon's automatic speech recognition software committed significantly more errors for black speakers than white speakers.

From Reuters
View Full Article

 

Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA


 

No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account