<![CDATA[Opaque artificial intelligence (AI) recommendations may undermine learners' ability to evaluate feedback and make informed decisions. This study examined whether an explainable artificial intelligence (XAI) learning system was associated with higher academic confidence and decision-making quality than a comparable AI system without explanatory output during undergraduate physics learning. A quantitative quasi-experimental pretest-posttest control-group design involved 120 students (mean age = 20.3 years, SD = 1.4; 54.2% female), with 60 students analyzed in each condition across eight weeks and treatment implemented through three intact courses. Academic confidence was measured using an 18-item study-adapted Academic Confidence Scale-Short Form (alpha = .87), whereas decision-making quality was assessed using a 16-item Learning Task Decision-Making Inventory (alpha = .83). ANCOVA controlling for corresponding pretest scores showed higher posttest academic confidence in the XAI condition, F(1, 117) = 52.34, p < .001, partial eta squared = .309, and higher decision-making quality, F(1, 117) = 38.71, p < .001, partial eta squared = .249. The results support an association between the explanatory-interface condition and both learner outcomes, but they do not establish confidence calibration, deliberative processing, or the proposed transparency mechanism. Interpretation is constrained by course-level allocation across only three clusters, unmodeled course dependence, a single-institution sample, study-adapted questionnaire outcomes, and the absence of a manipulation check or explanation-fidelity evidence.]]>
Copyrights © 2026
