Certified recourse guarantee improves fairness of gradient-boosted trees
CERBO: Certified Recourse at a Requested Rate for Gradient-Boosted Trees
Machine LearningArtificial IntelligenceComputers and Society
Summary
When people get unfavorable decisions from automated systems, it helps if they can change their situation easily to reverse those decisions. Existing methods for gradient-boosted trees struggle to guarantee how often this is possible. The authors propose CERBO, a method that ensures a certain level of guaranteed affordable changes, called recourse, for individuals rejected by the model. It works efficiently with popular machine learning tools and improves the guarantee from previous methods across different datasets.
What this means in practice
- •For credit risk teams: Use CERBO to create credit models that guarantee affordable changes for denied applicants, improving fairness compliance without reducing accuracy.
- •For judicial risk assessors: Deploy CERBO-enhanced bail decision models that certify individuals can meet conditions to improve their chances of release.
Authors
Farhan Shahriyar Hossain, Taufikur Rahman Fuad, Md. Zarif Bin Hasnat, Md Abrar Jahin, Md Rizwan Parvez
Abstract
Recourse-aware learning trains models so that people who receive an adverse decision can reverse it with an affordable change. For gradient-boosted decision trees (GBDTs), the existing method, RABIT, exposes a single recourse weight whose effect on the recourse rate is abrupt and data-dependent, and its training bound cannot certify recourse for any rejected individual. We present Certified Recourse Boosting (CERBO), which takes a requested recourse rate and returns a GBDT with certified actions: it tracks exact additive effort along tree paths, trains against committed boxes (a family of bounds that contains RABIT's and certifies a within-budget action for each training sample) and adjusts the weight after every tree until the certified rate meets the request. A calibration step guarantees the rate on new individuals, and the certifier runs on any tree ensemble, including XGBoost and LightGBM, in milliseconds per person without a solver. In RABIT's own experimental protocol, at equal accuracy, CERBO raises the recourse ratio from 0.90 to 0.99 on FICO and from 0.64 to 0.77 on Bail. Under additive costs on six data sets, it raises the area under the accuracy-recourse frontier in all 12 settings, and in simulated deployments it guarantees recourse to 70% of rejected individuals in 11 of 12 settings, against 2 for RABIT.