Systematic Literature Review: Explainable AI Definitions and Challenges in Education

There is no agreed-upon definition of Explainable AI yet. The review identified 15 different definitions of XAI across the literature, with significant overlap between concepts like ethics, trustworthiness, and technical explanations. This lack of standardization creates confusion about what XAI actually means in practice, particularly in educational settings.

Seven major categories of challenges must be addressed for XAI to work in education. Beyond technical difficulties, the research identified ethical concerns, human-computer interaction problems, trustworthiness issues, and policy gaps as equally important barriers. This suggests that implementing XAI in schools requires solutions across multiple domains, not just technological fixes.

The gap between transparency and actual understanding is a critical concern. Making AI systems "transparent" (showing how they work) doesn't automatically make them trustworthy or ensure that educators and students can meaningfully understand and act on the explanations provided—a distinction that has important implications for responsible deployment in schools.

Policy and guidelines for XAI in education are currently underdeveloped. The identification of policy and guideline challenges as a distinct category suggests that educational institutions lack clear frameworks for when, how, and why to use explainable AI, leaving schools without adequate guidance for responsible implementation.