Let Users Decide: Navigating the Trade-offs between Costs and Robustness in Algorithmic Recourse

As machine learning (ML) models are increasingly being employed to make consequential decisions, there has been a growing interest in developing techniques which can provide recourse to affected individuals. Majority of these techniques provide recourse under the assumption that the affected individuals will implement the prescribed recourses \emph{exactly}. However, recourses often get implemented in a noisy and inconsistent manner due to a variety of reasons e.g., an individual who was asked to increase their salary by \$500 may get a promotion which comes with a raise of \$505. Motivated by this, we study the problem of recourse invalidation in the face of noisy human responses. More specifically, we theoretically and empirically analyze the behavior of state-of-the-art algorithms, and demonstrate that the recourses generated by these algorithms are very likely to be invalidated (i.e., result in negative outcomes) if small changes are made to them. We further propose a novel framework, EXPECTing noisy responses (\texttt{EXPECT}), which addresses the aforementioned problem by explicitly minimizing the probability of recourse invalidation in the face of noisy responses. Our framework can ensure that the resulting recourses are invalidated at most $r \%$ of the time, where $r$ is provided as input by the end user requesting recourse. By doing so, our framework provides end users with greater control in navigating the trade-offs between recourse costs and robustness to noisy responses. Experimental evaluation with multiple real world datasets demonstrates the efficacy of the proposed framework, and validates our theoretical findings.