Upstart was founded more than a decade ago to improve access to credit through the application of innovative technology and data science. We have made significant progress towards this goal—serving more than 3 million customers and facilitating more than $39 billion in loans as of July 2024—and along the way we’ve developed industry-leading techniques for detecting and avoiding unlawful bias on our platform.
Advanced underwriting systems are reducing the racial inequities that plague traditional lending systems. Lenders that assess only traditional and backward-looking variables lock historical biases into their decisions. Forward-looking underwriting models, however, collect and use nontraditional underwriting data to assess each applicant’s future potential. Machine learning models learn from repayment events in a virtuous cycle. When borrowers who would have been declined by a traditional lender take out a loan from an Upstart lending partner, their repayment events train our machine learning models for future applicants and help to drive more approvals for future applicants that would have been declined by a traditional lender. This process repeats itself, and over time the model becomes increasingly fair in its outcomes.
Comprehensive fairness testing is conducted on an ongoing basis to statistically measure equity, accuracy, and inclusion across underwriting decisions and pricing outcomes on our platform.
Upstart fairness testing includes a robust analysis of disparate treatment and disparate outcome risk along with a less discriminatory alternative model search.
The results of these tests are shared with each lending partner to provide continued transparency into the integrity of Upstart’s models from a fair lending standpoint.
Upstart uses advanced modeling to improve our lending partners’ ability to underwrite accurately within the lending parameters set by the partner. The use of AI in credit underwriting offers tremendous benefit to consumers by more accurately assessing their ability to repay debts, increasing access to credit for thin-file consumers, and eliminating unconscious (or overt) bias in discretionary pricing and exception processes.
The consumers that are underwritten using AI models and the lenders that use AI models in their underwriting process must be able to interpret and explain the reason for the outcome.
Upstart has a proven methodology to explain AI model outcomes inclusive of approval/denial and pricing outcomes. Upstart’s industry-leading approach to explaining the outcomes of our models satisfies all legal and regulatory requirements and equips lending partners using the models to interpret and understand them.
Lenders can view the top reasons for approval and the top reasons for denial of any selected loan application in their lender dashboard.