A Comprehensive Guide to Interpretable AI-Powered Discoveries in Astronomy

The exponential growth of astronomical data necessitates the adoption of artificial intelligence (AI) and machine learning for timely and efficient scientific discovery. While AI techniques have achieved significant successes across diverse astronomical domains, their inherent complexity often obscures the reasoning behind their predictions, hindering scientific trust and verification. This review addresses the crucial need for interpretability in AI-powered astronomy. We survey key applications where AI is making significant impacts and review the foundational concepts of transparency, interpretability, and explainability. A comprehensive overview of various interpretable machine learning methods is presented, detailing their mechanisms, applications in astronomy, and associated challenges. Given that no single method offers a complete understanding, we emphasize the importance of employing a suite of techniques to build robust interpretations. We argue that prioritizing interpretability is essential for validating results, guarding against biases, understanding model limitations, and ultimately enhancing the scientific value of AI in astronomy. Building trustworthy AI through explainable methods is fundamental to advancing our understanding of the universe.