US Congress takes another run at AI accountability

An unsuccessful 2019 bill designed to make biased commercial AI algorithms [less biased] has been updated and re-introduced to the U.S. Congress.

The Algorithmic Accountability Act takes aim at “secret algorithms” that unfairly open or close opportunities to targeted communities, according to a statement published by Senator Ron Wyden, D-Ore. Wyden introduced the legislation along with Sen. Cory Booker, D-N.J., and Rep. Yvette Clarke, D-N.Y.,

Wyden and Booker, who is Black, drew direct parallels between biased code and crooked business practices that used to be more prevalent and accepted until the 1970s.

They mentioned steering homebuyers of color away from some White communities, but also commercial services that put at a disadvantage people of color and women. The same practices were used to discriminate against Jewish and non-binary consumers.

Energy for passage is being drawn from politically and philosophically motivated attacks on the country’s largest technology companies, including Google, Facebook and Amazon. A growing rejection of indiscriminate facial recognition surveillance also is fueling legislative efforts.

The Algorithmic Accountability Act would force firms to assess the negative impacts of “automating critical decision-making” in new and previously written code. Developers and users would have to assess and report findings.

It would enlist the Federal Trade Commission in regulating assessment and reporting back to it. Funding to hire 50 staff members and to create an enforcement division called the Bureau of Technology.

The commission would also publish an “anonymized aggregate report on trends.” It would create a public database listing decisions that have been automated, data sources, broad metrics and avenues for contesting decisions.

A handful of governments have implemented accountability policies, including New York City, Amsterdam, the Indian state of Tamil Nadu, the United Kingdom, Finland, Sweden and New Zealand.

Co-sponsors of the legislation, include Sens. Brian Schatz, D-Hawaii; Mazie Hirono, D-Hawaii; Ben Ray Luján, D-N.M.; Tammy Baldwin, D-Wis.; Bob Casey, D-Pa.; and Martin Heinrich, D-N.M.

Those endorsing the act include Accountable Tech, Aerica Shimizu Banks, Center for Democracy and Technology, Color of Change, Consumer Reports, Credo AI, the Electronic Privacy Information Center, Fight for the Future, IEEE, JustFix, Montreal AI Ethics Institute, OpenMined and Parity AI.

Article Topics

accuracy  |  AI  |  algorithmic accountability  |  algorithms  |  biometric bias  |  biometrics  |  biometrics research  |  facial recognition  |  legislation  |  research and development  |  United States