XAI-P-T: A Brief Review of Explainable Artificial Intelligence from Practice to Theory

Black-box AI systems require explanation methods because their decision-making processes are often opaque. The paper identifies categories of explanation approaches for these systems, emphasizing that as AI becomes more prevalent in high-stakes decisions, understanding why an AI system made a particular choice is increasingly important for accountability and trust.

Explainable AI draws from multiple academic disciplines, not just computer science. The paper notes that explanation has theoretical grounding across multidisciplinary fields, meaning responsible AI requires insights from philosophy, psychology, social science, and other domains to create explanations that are actually meaningful to humans.

There's a gap between XAI research in practice and theory that needs bridging. The authors identify that most existing literature focuses on either practical applications or theoretical foundations separately, but connecting these two perspectives is essential for helping both new researchers and experienced scholars develop more effective and responsible AI systems.

The field of XAI is still developing with multiple directions for future work. Rather than presenting a settled consensus, the paper indicates that explainable AI remains an evolving area where researchers are still determining best practices for making AI systems transparent and trustworthy.