A Survey of Algorithmic Recourse: Contrastive Explanations and Consequential Recommendations

Abstract

Machine learning is increasingly used to inform decision-making in sensitive situations where decisions have consequential effects on individuals’ lives. In these settings, in addition to requiring models to be accurate and robust, socially relevant values such as fairness, privacy, accountability, and explainability play an important role in the adoption and impact of said technologies. In this work, we focus on algorithmic recourse , which is concerned with providing explanations and recommendations to individuals who are unfavorably treated by automated decision-making systems. We first perform an extensive literature review, and align the efforts of many authors by presenting unified definitions , formulations , and solutions to recourse. Then, we provide an overview of the prospective research directions towards which the community may engage, challenging existing assumptions and making explicit connections to other ethical challenges such as security, privacy, and fairness.

PreCoF: counterfactual explanations for fairness

“…Several works list actionability and plausibility (adherence to data manifold) as desirable properties of counterfactual explanations (Guidotti, 2022;Karimi et al, 2021;Verma et al, 2020Verma et al, , 2021. These are two distinct concepts where the former restricts actions to those that are possible to do, and the latter requires that the resulting counterfactual instance is realistic or in line with the data manifold (Karimi et al, 2021).…”

“…The literature on algorithmic recourse has focused on finding "an actionable set of changes a person can undertake in order to improve their outcome" (Joshi et al, 2019;Karimi et al, 2021). Algorithmic recourse poses its own fairness criteria, where the effort to reach the required outcome is taken into account.…”

Directive Explanations for Monitoring the Risk of Diabetes Onset: Introducing Directive Data-Centric Explanations and Combinations to Support What-If Explorations

“…Adadi and Berrada [2] have defined counterfactual explanations as examplebased methods that provide minimum conditions required to obtain an alternate decision. Although counterfactuals provide useful model-agnostic post-hoc explanations, examples generated by counterfactual algorithms can be practically infeasible, contradictory, or uncontrolled, thereby indicating a need for actionable recourse [28,51]. For instance, to obtain a lower risk of diabetes, counterfactual algorithms can indicate patients to reduce their age by 30 years or alter their gender, which is practically infeasible.…”

Supervised Feature Compression based on Counterfactual Analysis

“…Alternatively, the term local explainability denotes the set of techniques that can be used to explain a single decision of a black-box model; among these methods, Counterfactual Explanations have been gaining an increasing popularity in recent years. Indeed, Counterfactual Explanations allow providing feedback to users on how to change their features in order to change the outcome of the decision (Karimi et al, 2021;Guidotti, 2022).…”

“…Counterfactual Explanations are a post-hoc local explainability technique (Karimi et al, 2021;Molnar et al, 2020). For an individual who has been subject to algorithmic decision making, and that has received an undesired decision, counterfactual analysis provides feedback on how to change the features of the individual in order to change the decision.…”