Searching for actual causes: Approximate algorithms with adjustable precision

2025-07-10

AI Summary

Existing XAI and causal inference methods primarily identify *potential causes*—factors that could influence an outcome—whereas users require *actual causes*: specific, factual conditions that genuinely produced a particular observed outcome. However, actual causality lacks a formal definition for non-Boolean, black-box, and stochastic systems, and its identification is NP-complete with no practical approximation algorithms. Method: We propose the first framework for actual causal discovery tailored to such systems. It integrates structural causal models with adaptive approximate search and introduces a polynomial-time algorithm with theoretical guarantees, enabling adjustable trade-offs between precision and completeness. Contribution/Results: Our approach overcomes dual limitations of conventional XAI—expressive inadequacy and computational intractability. Experiments demonstrate its ability to identify actual causes in complex systems where prior methods fail, while explanation accuracy and coverage scale gracefully with available computational resources.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Humans and AI: Explainable AI (XAI) for Human Understanding

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs

Abstract

Causality has gained popularity in recent years. It has helped improve the performance, reliability, and interpretability of machine learning models. However, recent literature on explainable artificial intelligence (XAI) has faced criticism. The classical XAI and causality literature focuses on understanding which factors contribute to which consequences. While such knowledge is valuable for researchers and engineers, it is not what non-expert users expect as explanations. Instead, these users often await facts that cause the target consequences, i.e., actual causes. Formalizing this notion is still an open problem. Additionally, identifying actual causes is reportedly an NP-complete problem, and there are too few practical solutions to approximate formal definitions. We propose a set of algorithms to identify actual causes with a polynomial complexity and an adjustable level of precision and exhaustiveness. Our experiments indicate that the algorithms (1) identify causes for different categories of systems that are not handled by existing approaches (i.e., non-boolean, black-box, and stochastic systems), (2) can be adjusted to gain more precision and exhaustiveness with more computation time.

Problem

• Formalizing actual causes in explainable AI remains unresolved
• Identifying actual causes is NP-complete with few solutions
• Proposing polynomial algorithms for adjustable precision cause identification

Innovation

• Polynomial complexity algorithms for causality
• Adjustable precision in cause identification
• Handles non-boolean, black-box, stochastic systems