[論文レビュー] The role of causality in explainable artificial intelligence

Gianluca Carloni, Andrea Berti+1 more|arXiv (Cornell University)|Sep 18, 2023

Explainable Artificial Intelligence (XAI)被引用数 9

ひとこと要約

A comprehensive review that maps how causality and XAI relate, identifying three main perspectives (critics to XAI, XAI for causality, causality for XAI) and outlining methodological and practical bridges between the fields.

ABSTRACT

Causality and eXplainable Artificial Intelligence (XAI) have developed as separate fields in computer science, even though the underlying concepts of causation and explanation share common ancient roots. This is further enforced by the lack of review works jointly covering these two fields. In this paper, we investigate the literature to try to understand how and to what extent causality and XAI are intertwined. More precisely, we seek to uncover what kinds of relationships exist between the two concepts and how one can benefit from them, for instance, in building trust in AI systems. As a result, three main perspectives are identified. In the first one, the lack of causality is seen as one of the major limitations of current AI and XAI approaches, and the "optimal" form of explanations is investigated. The second is a pragmatic perspective and considers XAI as a tool to foster scientific exploration for causal inquiry, via the identification of pursue-worthy experimental manipulations. Finally, the third perspective supports the idea that causality is propaedeutic to XAI in three possible manners: exploiting concepts borrowed from causality to support or improve XAI, utilizing counterfactuals for explainability, and considering accessing a causal model as explaining itself. To complement our analysis, we also provide relevant software solutions used to automate causal tasks. We believe our work provides a unified view of the two fields of causality and XAI by highlighting potential domain bridges and uncovering possible limitations.

研究の動機と目的

• Survey the interdisciplinary literature to understand how causality and XAI are intertwined.
• Identify the main perspectives on the causality–XAI relationship.
• Provide a unified view highlighting domain bridges and limitations between causality and XAI.

提案手法

• Structured literature review across Scopus, IEEE Xplore, Web of Science, and ACM databases.
• Keyword co-occurrence analysis using VOSviewer to map relationships among concepts.
• Topic clustering to organize papers into coherent perspectives: critics to XAI under a causal lens, XAI for causality, and causality for XAI.
• Compilation and analysis of software solutions cited for automated causal tasks.

実験結果

リサーチクエスチョン

• RQ1 What are the relationships between causality and XAI in the current literature?
• RQ2 What frameworks or formalisms bridge causality and XAI, and what are their limitations?
• RQ3 Which perspective (critics, XAI for causality, or causality for XAI) best supports integrating causality with XAI?
• RQ4 How can causal tools and metrics be applied to enhance XAI explanations?

主な発見

• Three main perspectives emerge: critics to XAI under a causal lens, XAI for causality, and causality for XAI.
• Lack of causality in current AI/XAI is highlighted as a major limitation, with concerns about robustness and trust in explanations.
• XAI can aid causal inquiry by suggesting pursue-worthy experimental manipulations, but explanations may be imperfect guides to causal inference.
• Causality concepts (SCMs, do-operator, causal Shapley values) can augment XAI methods and provide more mechanistic explanations.
• Access to causal models or causal reasoning can intrinsically explain AI decisions in critical settings.
• The review maps software solutions used to automate causal tasks and discusses their attributes.