Explainable AI for Failure Root Cause Analysis in Self-Adaptive Systems

Current ML approaches to testing autonomous systems, like those mentioned by Pezzè (2022), excel at detecting failures but struggle to explain them. While papers like Vashishtha & Sharma (2025) and Panda et al. (2024) show ML's effectiveness in prediction, they don't address the "why" question - crucial for debugging complex, self-adaptive systems. This research proposes using explainable AI techniques (like SHAP values or counterfactual explanations) to create causal maps between environmental conditions, system adaptations, and failures. For example, instead of just knowing that a self-driving car failed in rainy conditions, we'd understand that it specifically failed when rain intensity exceeded threshold X while the sensor fusion algorithm was in mode Y. This builds on the testing automation work by Nagila et al. (2025) but adds an explanatory layer. The innovation lies in treating failure explanations as first-class artifacts in the testing process, enabling developers to understand not just that the system failed, but why the adaptation logic made the wrong decision. This could revolutionize debugging for the kinds of systems Pezzè (2022) identifies as critical future challenges.

References:

1. Machine learning and natural language processing for automating software testing (tutorial). M. Pezzè (2022). ESEC/SIGSOFT FSE.
2. Estimation of Software Reliability Testing Using Machine Learning Techniques. Aditi Vashishtha, Shiv Kumar Sharma (2025). International Conference on Contemporary Computing.
3. Automating Fault Prediction in Software Testing Using Machine Learning Techniques: A Real-World Applications. Prasanta Panda, Debaryaan Sahoo, Debarjun Sahoo (2024). 2024 2nd International Conference on Sustainable Computing and Smart Systems (ICSCSS).
4. A Framework for Automated Software Testing using Machine Learning and Artificial Intelligence. Ashish Nagila, Neelu Trivedi, Ritu Nagila, Kanishk Trivedi, Sanjeev Bhardwaj, Jeetu Rani (2025). 2025 International Conference on Knowledge Engineering and Communication Systems (ICKECS).