User Experience Optimization in AI-Driven Learning Systems

Abstract

AI-driven learning systems -- adaptive tutors, recommendation engines, automated feedback generators, and predictive analytics dashboards -- introduce unique user experience challenges beyond conventional edtech UX: learners must understand and trust AI recommendations they cannot inspect; instructors must interpret AI analytics dashboards they did not design; and administrators must configure AI parameters they do not fully understand. The explainability, transparency, and controllability of AI system outputs become as important as prediction accuracy for user experience quality in AI-learning contexts. This paper proposes the AI Learning System UX Optimisation Framework (ALSUXOF), a systematic evaluation of six AI-specific UX dimensions -- AI transparency, recommendation explainability, human-AI control balance, AI feedback quality, trust calibration, and AI error recovery -- across 16 AI-driven learning systems evaluated by 1,920 users (learners, instructors, and administrators) in eight educational institutions over 12 months. ALSUXOF introduces the AI-UX Score (AIXS) and demonstrates that high-AIXS systems achieve 34.2% higher user adoption rates, 28.4% stronger AI recommendation acceptance, and 42.4% lower AI-related support tickets. Key results: recommendation explainability achieves the highest AIXS improvement impact (0.924 with SHAP explanations vs. 0.684 without); trust calibration is the strongest predictor of long-term AI system adoption; human-AI control balance is the most critical dimension for instructor acceptance. The framework provides AI-UX design guidelines for educational AI product teams.