A few governments write accountability policies to guide AI, biometrics
In an encouraging development, government officials are crafting accountability policies for the creation and use of algorithms, including those that perform biometric recognition.
The number of examples of this is small. “More than 40” policies in different stages of implementation in about 20 local and national governments were found for a new report written by a trio of non-profit organizations with a stake in the topic.
Publishing the work were the Ada Lovelace Institute, AI Now Institute and Open Government Partnership. Lovelace studies how AI should work for society. AI Now seeks social implications of algorithms and Open Government lobbies for transparency.
The three also found that the majority of policy writers reside in developed economies even though biometrics use and algorithmic decision making by government are global trends. And none are more than three years old.
Nonetheless, it would a safe bet that, on a percentage basis, public-sector bodies are more active than businesses is in holding systems (including framers, developers and users) accountable for AI-derived results.
The report makes a number of recommendations about design and implementation focused on policy monitoring and evaluation. According to the report’s authors, the field is too new to draw conclusions about what impact policies have or how effective they are.
Policymakers should integrate several practices in these early days, including real transparency about the assumptions in and objectives of accountability policy mechanisms.
Naturally, the authors suggest strategic engagement with agencies involved with AI as well as ongoing engagement with all communities impacted by biometrics and algorithm use.
The U.S. federal government touches on accountability policies infrequently and almost always in the form of a watchdog report warning of easily avoided missteps for policymakers.