Exploring LLMs for Causal Discovery in Cyber-Physical Systems

Causality plays a fundamental role in both human reasoning and complex system analysis. As Cyber-Physical<br/>Systems (CPS) become increasingly complex, understanding causal relationships between system events is<br/>essential for tasks such as anomaly detection and fault diagnosis. This paper explores the potential of Large<br/>Language Models (LLMs) to support causal discovery in CPS. In particular, the capabilities of LLMs in assisting<br/>domain experts to identify system states and their causal relations are investigated. We propose a hybrid workflow<br/>that integrates LLM-generated suggestions with domain expert validation, aiming to improve the efficiency of<br/>causal analysis. Our evaluation is conducted on a real-world smart grid use case and compares LLM-generated<br/>causal relations with domain expert-validated ground truth. The results indicate that, while LLMs can propose<br/>relevant causal structures, their effectiveness varies depending on the complexity of temporal and topological<br/>relationships between system states. Although these models do not replace human domain expertise, they can<br/>serve as a valuable tool for supporting causal discovery in a hybrid workflow. Future research should focus on<br/>refining LLM capabilities and expanding their application across different CPS domains. Investigating different<br/>LLMs, causal models, and larger datasets may provide deeper insights into their potential for causal discovery.