The paper discusses the development of a method to generate counterfactual explanations using answer set programming (ASP) and the s(CASP) system. The technique is designed to provide transparency for machine learning models, particularly those used in critical decision-making processes such as loan approvals, hiring etc. The generated explanations can help individuals understand why a certain decision was made and what changes could lead to a different outcome. The authors demonstrate how counterfactual explanations can be computed and justified, and how their algorithm can find the Craig Interpolant for a class of ASPs for a failing query.
Publication date: 23 Oct 2023
Project Page: https://arxiv.org/abs/2310.14497v1
Paper: https://arxiv.org/pdf/2310.14497
