Zest raises $ 15 million to reduce loan algorithm bias
Where is your business on the AI adoption curve? Take our AI survey discover.
Zest AI, a company developing AI-powered lending decision products, today closed a $ 15 million funding round led by Insight Partners. A spokesperson said the capital would be used to speed up Zest’s product marketing and R&D efforts.
About 1 in 9 (10.8%) loan applications for the purchase of a home – and more than 1 in 4 (26.4%) for refinancing – were refused in 2017, according to a national analysis of lender data conducted by the United States Bureau of Consumer. Financial protection. Minorities were disproportionately rejected, with the overall denial rate of black American mortgage applications reaching 18.4% in 2018 (Hispanic and Asian applicants were rejected 13.5% and 10.6% respectively of the time, compared to 8.8% for non-Hispanic white applicants.)
Zest, who was co-founded in 2009 by former Google IT director Douglas Merrill and former Sears vice president Shawn Budde, says its mission is to create “higher” standards around debugging algorithmic lending. To this end, the company helps banks, credit unions and specialty lenders identify borrowers by looking at more than just credit scores. Zest claims that institutions that lend using its models, including Discover, Akbank and VyStar, have seen a 20% increase in approval rates on average and a reduction of up to 50% in delinquencies, or debts. statements that an amount of debt is unlikely.
Zest provides more than 30 clients with resources to prepare, create, iterate and document machine learning decision models for cards, auto loans, personal loans, mortgages and student loans. Complementary tools help teams assess and validate models for security, stability, business impact, and compliance. Customers can deploy and monitor algorithms in production, or they can hire Zest’s team of service and machine learning experts to help with development and validation.
Zest claims to use a technique called contradictory debuffing to minimize the potential bias of the model. The technique pits two machine learning models against each other, one attempting to predict creditworthiness while the other guesses the candidate’s race, gender, and other attributes noted by the first model. Competition pushes the two to improve their methods until the predictor can no longer distinguish the outputs of race or gender from the first model, resulting in a model that is conspicuously more accurate and fair.
Zest recently introduced ZAML Fair, which the company says can reduce bias in loan portfolios with “little or no” impact on profitability. ZAML Fair leverages the transparency tools built into Zest’s suite of solutions to rank variables in a system based on the extent to which they lead to biased results. He then attempts to attenuate the influence of these signals to produce a superior model.
Based on mortgage lenders who have tested ZAML Fair, Zest says the tool would eliminate 70% of the national gap between Hispanic and White applicant approval rates and narrow the gap even greater between black and white borrowers. whites of more than 40%. In one blog postt, Zest cited a Harris Poll survey that found that a majority of Americans would give up more personal data if it resulted in a fairer credit decision. With that in mind, Zest believes he can reduce bias by using “better math and more data to assess borrowers.”
Of course, it is difficult, if not impossible, to completely rid algorithms of bias. Facial recognition models to ignore Blacks, Middle Easterners, and Latinxes more often than those with fairer skin. AI researchers at MIT, Intel, and the Canadian AI initiative CIFAR have found high levels of bias on the part of some of the most popular pre-trained models. And algorithms developed by Facebook was found to be 50% more likely to deactivate black user accounts compared to white users.
But Zest says the data proves his efforts are making a difference. Using Zest’s underwriting software platform, lender claims it was able to reduce the disparity in approval rates between white applicants and applicants of color by 30% on average, without increasing risk of the wallet. Separately, an auto lender was able to approve “thousands” of additional borrowers.
“The COVID-19 shock has led many financial institutions to update and improve their systems for resilience and sustainability, which has resulted in a significant increase in demand for our business,” said CEO Mike de Vere to VentureBeat by email. “Much of this included building new and improved subscription systems with the latest math and software technologies. This resulted in Zest’s best Q2 ever, with a view to ending the year with triple-digit growth.
Los Angeles-based Zest has raised more than $ 87 million in venture capital to date.
VentureBeat’s mission is to be a digital public place for technical decision-makers to learn about transformative technology and conduct transactions. Our site provides essential information on data technologies and strategies to guide you in managing your organizations. We invite you to become a member of our community, to access:
- up-to-date information on the topics that interest you
- our newsletters
- Closed thought leader content and discounted access to our popular events, such as Transform 2021: Learn more
- networking features, and more