Explaining Groups of Instances Counterfactually for XAI: A Use Case, Algorithm and User Study for Group-Counterfactuals

Greta WarrenMark T. KeaneChristophe GueretEoin Delaney

Mar 2023

摘要

Counterfactual explanations are an increasingly popular form of post hocexplanation due to their (i) applicability across problem domains, (ii)proposed legal compliance (e.g., with GDPR), and (iii) reliance on thecontrastive nature of human explanation. Although counterfactual explanationsare normally used to explain individual predictive-instances, we explore anovel use case in which groups of similar instances are explained in acollective fashion using ``group counterfactuals'' (e.g., to highlight arepeating pattern of illness in a group of patients). These groupcounterfactuals meet a human preference for coherent, broad explanationscovering multiple events/instances. A novel, group-counterfactual algorithm isproposed to generate high-coverage explanations that are faithful to theto-be-explained model. This explanation strategy is also evaluated in a large,controlled user study (N=207), using objective (i.e., accuracy) and subjective(i.e., confidence, explanation satisfaction, and trust) psychological measures.The results show that group counterfactuals elicit modest but definiteimprovements in people's understanding of an AI system. The implications ofthese findings for counterfactual methods and for XAI are discussed.

展开全部

图表提取

论文十问

参考文献

被引用

提问与回答

暂无人提供速读十问回答

被引用

笔记

问答

领域