A Human-Centric Approach to Explainable AI for Personalized Education

Explainability is a critical barrier to AI adoption in education: The paper identifies that despite AI's strong predictive performance, teachers, students, and parents lack trust in these systems due to inability to understand how AI makes decisions—a gap that prevents real-world classroom implementation of personalized learning tools.

Post-hoc explanations may be unreliable: The research reveals "systematic disagreements between post-hoc explainers," suggesting that common methods of explaining AI decisions after the fact may not be trustworthy, highlighting the need for AI systems designed to be interpretable from the ground up rather than explained retroactively.

Human needs should drive AI design, not just technical performance: The paper emphasizes a "human-centric approach" by conducting human studies alongside technical development, demonstrating that understanding what students, teachers, and parents actually need from explanations is as important as building accurate AI models.

Different stakeholders need different types of explanations: By focusing on personalized education, the research acknowledges that students, teachers, and parents have distinct needs—suggesting that responsible AI requires tailoring explanations to specific user groups rather than applying one-size-fits-all approaches.

Interpretable AI architectures should be built into systems from the start: The paper proposes novel "inherently interpretable model architectures" rather than relying solely on explanation tools added afterward, suggesting that trustworthy AI for high-stakes domains like education requires fundamental redesign of how AI systems are constructed.