A Comprehensive Study on Adaptive Learning and Assessment with Explainable AI

Adaptive learning systems are a cornerstone of contemporary educational practice and develop personalized content and assessments based on the individual learner profile. However, traditional AI-based learning experiences, particularly those driven by machine learning algorithms, do not communicate the reasoning behind the system's recommendations for learning pathways, which can limit the user’s trust, transparency, and interpretability of feedback. This study proposed a scalable and comprehensive adaptive learning and assessment framework that incorporated Explainable Artificial Intelligence (XAI) as an overarching mechanism that enhances the usefulness of decision making and learning outcomes. The methodology encompassed data collection (learner data), creating personalized learning pathways using AI algorithms, automating an assessment's grading process, completing an assessment with XAI functions generating interpretable feedback, and finally satisfying to the Centre for Education Statistics and Evaluation (2018) authenticity criterion. A comparative evaluation was undertaken against a rule-based, non-XAI model and a manual approach to grading the assessments. The comparative evaluation was based on a framework which allowed for contextualization against established variables (e.g. personalization accuracy, assessment automation efficiency, interpretability of feedback, user trust, etc.). The data showed the AI + XAI model out-performed all baselines; 92% personalization accuracy, 95% assessment automation efficiency, and the highest percentage scores for interpretability and user trust. Consequently, using the approaches presented in the study improved existing AI-based methods and added meaningful learner transparency about the use of explainable AI - specifically, how explainable approaches could increase learner confidence in adaptive learning and assessment designed with AI technology in the field of education.