Font Size: 

A study carried out on a dataset and a mathematical model to supportdecision making in the credit process for established clients in a commercial bankin Costa Rica will be presented. The main objective of the study was to evaluatealternatives to mitigate the gender biases present in the model. To achieve this,possible sources of bias in the modelwere identified, among which possible disparatetreatment, association, selection, malicious, and automation biases were identified.These biases were then measured in more detail, finding that they are small, exceptperhaps for the selection bias. Thirdly, alternative models were built to mitigatethese biases, to finally evaluate the difference both in the fairness measures that wereused and in the performance of the alternative models compared to the original todetermine the one that provides greater value to the business. Here, it was foundthat the gains are minor and that what could be more worthwhile is to maintain thecurrent model and investigate other credit scoring models used in other stages of thecredit granting process.

gender bias, bias mitigation, credit scoring