Explainable AI in Adaptive Assessments: Building Trust and Transparency in Continuous Learning

As artificial intelligence becomes central to adaptive assessment systems, the opacity of the underlying models is starting to threaten something fundamental: the trust, fairness, and pedagogical value of automated decisions. Explainable AI (XAI) offers a way through this, making the reasoning behind adaptive difficulty selection, automated scoring, and personalized feedback transparent and interpretable to learners, educators, and institutions alike. In this paper, I look at how XAI techniques, including feature attribution, counterfactual explanation, rule extraction, and interpretable skill modeling, can be embedded into AI-driven adaptive assessments to strengthen continuous learning and accessibility. Building on prior work in reinforcement learning for personalized learning paths, I explore how explanations turn opaque recommendations into guidance learners can actually act on. Along the way, I examine how transparency builds learner trust, helps mitigate algorithmic bias, manages cognitive load, and meets emerging regulatory requirements. Through illustrative case studies and comparative tables, I show that explainability improves engagement, supports fairer outcomes, and gives educators the means to audit and refine their assessment models. I close by outlining the benefits, challenges, and open research directions for deploying trustworthy, interpretable AI in education, arguing that explainability isn't an optional add-on; it's a foundational requirement for responsible adaptive assessment at scale.