Abstract: Explainable AI (XAI) is becoming increasingly important as machine learning systems are deployed in sensitive domains where trust, transparency, and accountability are essential. While many current approaches focus on post-hoc explanations, causality offers a deeper and more principled way to understand model behavior.
In this talk, Dr. Bapi Chatterjee and Dr. Vineeth N. Balasubramanian will explore how causal reasoning can strengthen explainability in AI systems. They will discuss the motivation for integrating causal inference into XAI, highlighting its potential to uncover hidden biases, improve fairness, and provide more robust insights into decision-making processes.
Key themes will include:
The limitations of correlation-based explanations and the need for causal models.
Methods for embedding causal structures into machine learning pipelines.
Case studies demonstrating how causality enhances transparency and reliability.
Future directions for explainable AI that balance interpretability with performance.
The session will provide both theoretical foundations and practical examples, showing how causality can transform explainable AI from descriptive tools into frameworks that genuinely improve trust and usability in real-world applications.