CFPB says AI is not an excuse for vague or unclear explanations for credit rejections

The news: The Consumer Financial Protection Bureau reminded financial institutions (FIs) that if they deny credit credit to consumers, they must still explain why, accurately and in detail—even if AI or an algorithm made the decision.

Give me good reasons: A CFPB press release warns FIs not to use generic boilerplate or checklists in their adverse action notices.

• To comply with the Equal Opportunity Credit Act, FIs that deny credit applications must send out an adverse action notice explaining why the consumer was rejected.
• The notices must show real, data-driven reasons for the rejection and provide evidence that the consumer wasn’t discriminated against for race, religion, or other protected status.
• For instance, rather than attributing a restriction on a consumer’s line of credit to their “purchasing history,” the notice needs to indicate which purchases or stores they shopped at hurt their credit.

The big picture: FIs are increasingly turning over credit decisions to AI and complex algorithms which don’t use the same logic as traditional computer programs in the underwriting process.

• Instead, these models process massive amounts of data and “learn” from it, resulting in highly complex internal workings.
• The lack of explanation for these workings or conclusions makes it a “black box” that’s opaque and hard to see into, sometimes even for programmers and engineers.
• The CFPB has previously reminded consumers that federal law entitles them to an explanation for credit denials—and that right doesn’t disappear if the FI uses an algorithm so complex, it doesn’t really understand why.

The memorable quote: CFPB Director Rohit Chopra said in a statement, “Creditors must be able to explain the reason for their denial. There is no exemption for artificial intelligence.”

Our take: The CFPB’s warning is another example of the disconnect between consumers and companies that are racing to take advantage of the efficiencies that AI products offer.

• As Insider Intelligence Analyst Daniel Konstantinovic writes, “Consumers haven’t yet warmed up to AI, and a top concern is how AI’s constant learning technology will affect their privacy.”

Action item: Consumers may be surprised or upset to learn that their credit evaluations included data that may not intuitively relate to their finances. If their adverse action notices don’t offer clear explanations, FIs need to prepare scripts for customer-facing staff who may encounter these situations.