AI and Machine Learned-Based Credit Underwriting and Adverse Action Under the ECOA

By Eric Knight


In the rapidly evolving retail financial services market, new technologies, including artificial intelligence and machine learning, are challenging the premises of existing laws and revolutionizing the process of loan and credit underwriting. These technologies allow creditors to consider a wide range of alternative factors which are untapped by traditional credit scoring models, with research suggesting that they can allow companies to extend affordable credit to members of underserved communities who currently lack the credit history necessary to fully participate in the financial system. Using these tools, businesses can increase efficiency while potentially maintaining or lowering risk levels and delinquency rates.

Despite the potential benefits of these methods of credit decision making, they are difficult to fit neatly into the framework of existing fair lending laws like the Equal Credit Opportunity Act which, among other things, requires creditors to provide applicants with a statement of specific reasons for adverse action taken in connection with a credit application. These notices are intended to give consumers the information necessary to contest unfair credit decisions, dispute incorrect information in their credit report, and improve their creditworthiness for future transactions.

In a world in which credit decisions are based on a potentially vast and evolving set of factors, some of which have no intuitive relationship to creditworthiness, it is hard to see how these public policy goals can be achieved under existing frameworks. If they are forced to comply with existing law, creditors may be compelled to forego use of these new technologies, with limited countervailing benefits to consumers.

This Note seeks to resolve fundamental conflicts between existing law and these developing technologies. I advocate for a light regulatory touch to adverse action notices, with the goal of fostering innovative approaches to credit decisionmaking while remaining mindful of the consumer protection goals of the Equal Credit Opportunity Act. I will attempt to outline the basic principles of AI and machine learning-based credit scoring, how consumers and businesses alike stand to benefit from their implementation, how they fit (or do not fit) within the current legal regime, and a potential solution for this new frontier of credit decision making