Home » News » Bad Data, Poor Decisions, Even Worse Consequences: A Case for Transparency in Automated Lending Practices

Bad Data, Poor Decisions, Even Worse Consequences: A Case for Transparency in Automated Lending Practices

With the spotlight on CNN’s recent report on the disparities in lending by the largest credit union in the U.S., our work is reinforced as current systems continue to reveal the need for transparency and accountability in automated lending decisions.

The report found vast disparities; in particular, 77.1% of White applicants were approved compared to 48.5% of Black applicants, a 28.6 percentage-point gap. Even when applicants shared more than a dozen different variables, Black applicants were more than twice as likely to be denied. While we believe lenders don’t set out to build biased lending models, the clear discrepancies in their denial rates, shows the need for accountability.

By using solutions like UnBias™, this lender could have proactively identified the disparity in approval rates in addition to pinpointing the specific factors in the application process that contributed the most to this difference. Furthermore, the comparative disparate treatment testing within UnBias™ would have helped in recognizing whether discretionary choices made by underwriters demonstrated systematic bias, providing the opportunity to address this accordingly. Another point the report made was that there weren’t clear answers as to why folks were being denied, opening the conversation about transparent lending models.

Here at Stratyfy, we empower institutions to make confident decisions while having full insight on how and why decisions were made. This enables lenders to provide clear answers regarding the factors that determined the outcome. Additionally, in the case of a negative outcome, having the right answers allows lenders to build upon that customer relationship to help guide them to a more favorable one.

At a time where technology can play a pivotal role in ensuring fairness, why do we continue to rely on legacy systems that can both harm the people being served and, in this case, the organization’s reputation?