At ZestFinance, we strive to make fair and transparent credit available to everyone. We do that by providing lenders with superior AI software to create powerful credit models that uncover worthy borrowers who’ve been overlooked by status quo credit-scoring techniques.
Too many Americans struggle to get credit. Millions of good applicants get rejected for loans due to gaps, errors, or structural inequities built into their credit file. The deck is especially stacked against minority groups and young people. The disparity in home loan approval rates between whites and minority borrowers is a pernicious fact of life in the U.S. and contributes to our country’s stubborn wealth gap. There is room for improvement.
That’s why we’re launching ZAML Fair™, new AI software Zest customers can use to reduce or eliminate bias in their loan portfolios with little or no impact on profitability. ZAML Fair shrinks the racial disparity gap of almost any model. It's the latest extension to Zest’s flagship ZAML suite, which lenders around the world rely on to safely develop and deploy machine learning (ML) credit underwriting.
What Is ZAML Fair?
Today, most lenders assess their credit models for fairness only after they’ve built and trained the model. That limits their options to either tossing out offending variables and re-running the model (aka whack-a-mole) or making the case for why an offending variable is needed.
We don’t think banks should have to choose between fairness and performance. Machine learning can produce a model that’s both fairer and more accurate than the one they’re using today.
ZAML Fair builds on the success that our benchmark ZAML suite has already achieved in expanding economic opportunity. For one auto lender, ZAML increased the approval rate among millennials by 25% simply by using more data and better math to spot more good borrowers. ZAML also tripled the approval rate for “thin-file applicants” (people with fewer years of credit and often lower credit scores).
How ZAML Fair works
ZAML Fair is a new algorithm that lenders can apply to an existing credit model to tune down the impact of discriminatory credit data. ZAML Fair uses the transparency tools built into ZAML to rank a model's credit variables by how much they lead to biased outcomes, and then carefully mutes the influence of those signals to produce a better model in a fraction of the time and effort required by legacy techniques. No more whack-a-mole.
ZAML Fair works on linear models such as those produced by logistic regression as well as ML models including deep neural networks, random forests, and gradient-boosted decision trees.
Several mortgage lenders have tested ZAML Fair and, based on their results, ZAML Fair applied widely would eliminate 70% of the nation's gap in approval rates between Hispanic and white mortgage applicants, and cut the even larger gap between black and white borrowers by more than 40%. This would put more than 172,000 minority families into new homes. We're ready to help lenders make that happen.
ZAML Fair is available now for all ZAML customers. If you want to talk to one of our sales reps, click here.