Use of the tool can slash the gap in approval rates between white and minority borrowers, company says
AI software provider ZestFinance has unveiled ZAML Fair, a new software tool tested by mortgage lenders to reduce bias and discrimination in lending.
Before launching the tool, ZestFinance asked several mortgage lenders to test the ZAML Fair algorithm. The results revealed that ZAML Fair could cut 70% of the gap in US mortgage approval rates between Hispanic and white borrowers and remove over 40% of the bigger gap between black and white borrowers, the company said. This would help more than 172,000 minority families buy new homes, according to ZestFinance.
"The way lenders take bias out of their models today is by simply tossing out offending credit signals. That leaves a lot of performance on the table," said ZestFinance founder and CEO Douglas Merrill. "Banks shouldn't have to choose between fairness and accuracy, and ZAML Fair helps them optimize for both."
Lenders could utilize ZAML Fair as a new algorithm to tune traditional models for optimum fairness and automatically lessen the impact of discriminatory credit data, the company said.
“ZAML Fair uses the transparency tools built into ZAML to rank credit signals by how much they lead to biased outcomes and then produces a new model with maximum fairness and accuracy. ZAML Fair carefully reduces the influence of factors that drive disparity, including some common credit signals such as income and the traditional credit score. Lenders get a series of better models they can choose from at a fraction of the time and effort required by legacy techniques,” said ZestFinance in a statement.
ZAML Fair is an extension of ZestFinance’s flagship ZAML platform, which enabled lenders worldwide to safely develop and deploy machine learning (ML) credit underwriting.
