Software Engineering Principles for Fairer Systems: Experiments with GroupCART

2025-04-17

AI Summary

Traditional decision trees (e.g., CART) employ information gain with respect to the target variable as their splitting criterion, which often inadvertently introduces discriminatory bias along protected attributes such as gender or race. To address this, we propose GroupCART—a fairness-aware decision tree ensemble method intrinsically integrated into the modeling process. GroupCART explicitly incorporates maximization of entropy over protected attributes into its splitting criterion, jointly optimizing predictive accuracy and group-level fairness. It features a user-tunable accuracy–fairness trade-off parameter and requires neither preprocessing nor postprocessing. Built upon multi-objective entropy optimization and an extension of the CART framework, GroupCART preserves model interpretability and controllability. Empirical evaluation across multiple benchmark datasets demonstrates that GroupCART significantly improves fairness metrics—including statistical parity—while incurring less than a 2% drop in classification accuracy. The implementation and datasets are publicly available.

Technology Category

• Machine Learning: Ensemble Methods
• Computer Vision: Bias, Fairness & Privacy
• Philosophy and Ethics of AI: Bias, Fairness & Equity

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Fairness, privacy, and diversity in economic environments
• Social Networks and Social Media: Fairness and bias in social network and social media analysis

Abstract

Discrimination-aware classification aims to make accurate predictions while satisfying fairness constraints. Traditional decision tree learners typically optimize for information gain in the target attribute alone, which can result in models that unfairly discriminate against protected social groups (e.g., gender, ethnicity). Motivated by these shortcomings, we propose GroupCART, a tree-based ensemble optimizer that avoids bias during model construction by optimizing not only for decreased entropy in the target attribute but also for increased entropy in protected attributes. Our experiments show that GroupCART achieves fairer models without data transformation and with minimal performance degradation. Furthermore, the method supports customizable weighting, offering a smooth and flexible trade-off between predictive performance and fairness based on user requirements. These results demonstrate that algorithmic bias in decision tree models can be mitigated through multi-task, fairness-aware learning. All code and datasets used in this study are available at: https://github.com/anonymous12138/groupCART.

Problem

• Address discrimination in classification with fairness constraints
• Optimize decision trees to reduce bias in protected attributes
• Balance predictive performance and fairness in model construction

Innovation

• GroupCART optimizes entropy in target and protected attributes
• Achieves fairness without data transformation or major performance loss
• Customizable weighting balances predictive performance and fairness