Federal watchdog cautions about potential bias perpetuation in lending due to AI and machine learning

The utilization of artificial intelligence and machine learning in lending practices has recently caught the attention of the Federal Reserve’s top watchdog. In his speech at the National Fair Housing Alliance’s 2023 national conference, Michael Barr, the Fed’s vice chair of supervision, acknowledged the significant potential of these technologies but also highlighted the risk of perpetuating bias and disparities in lending. He explained that while AI tools can increase access to credit for more individuals at a relatively low cost, they can also amplify existing biases present in the training data or make inaccurate predictions.

To address discrimination in mortgage transactions, the Fed has introduced two policy initiatives. The first is a proposed rule to implement quality control standards in automated valuation models, which requires institutions involved in certain credit decisions to adopt policies and control systems that ensure credible estimates and protect against data manipulation. The second initiative involves seeking public comment on guidance to help financial institutions incorporate “reconsiderations of value” into their home appraisal process. By allowing for reassessment, the aim is to mitigate the risk of undervaluing real estate due to errors, omissions, or discrimination within the valuations.

Barr emphasized the importance of homeownership as a means for families to build wealth and expressed his full support for both policy initiatives. Additionally, federal agencies including the Federal Trade Commission and the civil rights division at the Department of Justice have committed to taking action against automated systems that result in harmful business practices. Barr’s remarks on fair lending garnered applause from the audience and were praised by Lisa Rice, the President and CEO of the NFHA, who lauded Barr for speaking out against algorithmic bias.