A Systematic Approach to Auditing and Mitigating Demographic Bias in Algorithmic Lending Models

Algorithmic credit-scoring and lending models, increasingly used to automate loan approval decisions, risk encoding and amplifying historical demographic biases present in training data, potentially resulting in systematically unfair outcomes for protected groups. This project develops a systematic auditing framework to detect demographic bias in algorithmic lending models across multiple fairness metrics, coupled with mitigation techniques applied at the pre-processing, in-processing, and post-processing stages, to produce lending models that are both predictive and demonstrably more equitable.