Performative Validity of Recourse Explanations

2025-06-18

AI Summary

This paper addresses the “performative failure” of counterfactual recourse in algorithmic loan denials—where recourse recommendations become invalid when many applicants act upon them, inducing distributional shifts and dynamic repositioning of the model’s decision boundary. Method: We formalize and analyze recourse validity under performative settings by integrating causal inference, game-theoretic modeling, and statistical learning theory. We propose a causally constrained recourse generation criterion that restricts interventions to causal variables only. Contribution/Results: We establish necessary and sufficient conditions for long-term recourse validity under model retraining, proving that interventions limited to causal variables guarantee robustness against performative effects. This challenges prevailing non-causal explanation paradigms and provides the first theoretical framework ensuring recourse validity in dynamic, strategic environments. Our results yield both foundational guarantees for trustworthy algorithmic decision-making and actionable design principles for deployable recourse systems.

Technology Category

• Reasoning under Uncertainty: Causality
• Machine Learning: Causal Learning
• Knowledge Representation and Reasoning: Action, Change, and Causality

Application Category

• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

When applicants get rejected by an algorithmic decision system, recourse explanations provide actionable suggestions for how to change their input features to get a positive evaluation. A crucial yet overlooked phenomenon is that recourse explanations are performative: When many applicants act according to their recommendations, their collective behavior may change statistical regularities in the data and, once the model is refitted, also the decision boundary. Consequently, the recourse algorithm may render its own recommendations invalid, such that applicants who make the effort of implementing their recommendations may be rejected again when they reapply. In this work, we formally characterize the conditions under which recourse explanations remain valid under performativity. A key finding is that recourse actions may become invalid if they are influenced by or if they intervene on non-causal variables. Based on our analysis, we caution against the use of standard counterfactual explanations and causal recourse methods, and instead advocate for recourse methods that recommend actions exclusively on causal variables.

Problem

• Examines validity of recourse explanations under performative effects
• Identifies conditions for recourse actions to remain valid
• Advocates causal-variable-focused recourse over standard methods

Innovation

• Formal conditions for valid recourse explanations
• Avoid non-causal variables in recourse actions
• Advocate causal-variable-exclusive recourse methods