CFPB, Federal Reserve Board: AI bias can't be ignored | Bank Innovation | Bank Innovation

CFPB, Federal Reserve Board: AI bias can’t be ignored

Andrew Harrer/Bloomberg

Following a recent backlash resulting from allegedly discriminatory underwriting methods for the Apple Card, lenders’ moves are facing scrutiny from regulators. In light of the increased oversight, however, industry practitioners emphasize the role of AI as a catalyst for a more inclusive process.

Albert Chang, counsel at the Consumer Financial Protection Bureau‘s innovation office, said discrimination allegations shouldn’t be taken lightly. While regulations should enable innovation, he argued that companies need to be mindful of the impact of biases in underwriting models.

“If you’re facing those allegations, the first step is to understand why the algorithm in the decision-making process led to two different outcomes for apparently similarly situated applicants,” he said, speaking at The Clearing House annual conference in New York on Thursday.

Meanwhile, Carol Evans, associate director of the division of consumer affairs at the Federal Reserve Board, said the fact that AI algorithms were developed by men creates opportunities for bias to creep into the process. As a result, she noted that diverse teams are crucial to combating bias.

“Diversity and inclusion matters in this discussion,” she said. “Who was at the table when the model was discussed?”

See also: Apple Card’s gender-bias claims look familiar to old-school banks

Lenders, however, expressed optimism about AI’s prospects as an enabler for change. Meredith  Fuchs, chief counsel of regulatory advisory at Capital One emphasized that institutions should consider AI-based underwriting as one component of a larger effort to align products with customer needs. To ensure the effectiveness and quality control of AI-based underwriting methods, humans need to be involved, she noted.

“You don’t just sort of unleash the machine to make a decision — humans are always involved in the decision making,” she said. “The machines allow humans to set criteria, to gather more data, to identify correlation [that people] couldn’t identify before, but we still set the outcome criteria,” she explained.

Annie Delgado, chief compliance officer at Upstart, expressed confidence that the technology will evolve. To Upstart, greater consistency of underwriting processes across the industry could help root out discrimination. 

“If you’re talking about a machine learning model and you have a more limited group of people that can control the inputs and monitor the outputs of that system, you can teach a machine to not have biases,” she said.

Bank Innovation Ignite, which will take place March 2-3 in Seattle, explores emerging technologies in banking. Attendees will gain a firsthand look at new products and services from both legacy and upstart perspectives. Request your invitation