Relevance-aware Algorithmic Recourse
Dongwhi Kim, Nuno Moniz
TL;DR
RAAR addresses the unrealistic assumption of equal-domain-value treatment in algorithmic recourse by incorporating a relevance function φ into a Bayesian-optimization framework for regression tasks. It uses a Gaussian-process surrogate and an Upper Confidence Bound acquisition to optimize both outcome changes and domain relevance, with two objective pathways: Y optimization and Relevance optimization. Across 15 datasets and two regression models, RAAR delivers recourses that are comparably effective to baselines while being more efficient and lower in cost, particularly under target-value scenarios. This approach enhances practical usefulness of counterfactual explanations, aligning recourse with user-domain priorities and potentially improving fairness in real-world decisions.
Abstract
As machine learning continues to gain prominence, transparency and explainability are increasingly critical. Without an understanding of these models, they can replicate and worsen human bias, adversely affecting marginalized communities. Algorithmic recourse emerges as a tool for clarifying decisions made by predictive models, providing actionable insights to alter outcomes. They answer, 'What do I have to change?' to achieve the desired result. Despite their importance, current algorithmic recourse methods treat all domain values equally, which is unrealistic in real-world settings. In this paper, we propose a novel framework, Relevance-Aware Algorithmic Recourse (RAAR), that leverages the concept of relevance in applying algorithmic recourse to regression tasks. We conducted multiple experiments on 15 datasets to outline how relevance influences recourses. Results show that relevance contributes algorithmic recourses comparable to well-known baselines, with greater efficiency and lower relative costs.