Automatic Failure Attribution and Critical Step Prediction Method for Multi-Agent Systems Based on Causal Inference

2025-09-10

AI Summary

Multi-agent systems (MAS) deployed in real-world settings suffer from poor fault attribution, as existing statistical correlation–based approaches fail to identify root-cause execution steps, resulting in low accuracy on standard benchmarks. To address this, we propose the first multi-granularity causal attribution framework tailored for MAS, integrating backward data-flow modeling, a novel CDC-MAS causal discovery algorithm, and the performance causal inversion principle—enabling root-cause localization and critical failure step prediction under non-stationary agent interactions. Our method unifies Shapley-value-based attribution, counterfactual simulation, and end-to-end causal structure learning to establish a diagnosis–optimization closed loop. Evaluated on the Who&When and TRAIL benchmarks, our approach achieves up to 36.2% step-level attribution accuracy and generates optimization suggestions that improve task success rates by an average of 22.4%.

Technology Category

• Multiagent Systems: Adversarial Agents
• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Multi-agent systems (MAS) are critical for automating complex tasks, yet their practical deployment is severely hampered by the challenge of failure attribution. Current diagnostic tools, which rely on statistical correlations, are fundamentally inadequate; on challenging benchmarks like Who&When, state-of-the-art methods achieve less than 15% accuracy in locating the root-cause step of a failure. To address this critical gap, we introduce the first failure attribution framework for MAS grounded in multi-granularity causal inference. Our approach makes two key technical contributions: (1) a performance causal inversion principle, which correctly models performance dependencies by reversing the data flow in execution logs, combined with Shapley values to accurately assign agent-level blame; (2) a novel causal discovery algorithm, CDC-MAS, that robustly identifies critical failure steps by tackling the non-stationary nature of MAS interaction data. The framework's attribution results directly fuel an automated optimization loop, generating targeted suggestions whose efficacy is validated via counterfactual simulations. Evaluations on the Who&When and TRAIL benchmarks demonstrate a significant leap in performance. Our method achieves up to 36.2% step-level accuracy. Crucially, the generated optimizations boost overall task success rates by an average of 22.4%. This work provides a principled and effective solution for debugging complex agent interactions, paving the way for more reliable and interpretable multi-agent systems.

Problem

• Attributing failures in multi-agent systems accurately
• Identifying critical failure steps using causal inference
• Improving task success rates through automated optimization

Innovation

• Causal inversion principle with Shapley values for agent blame
• CDC-MAS algorithm handles non-stationary interaction data
• Automated optimization loop using counterfactual simulation validation