Adaptive Learning Rate Optimization for Personalized Educational Interventions in Autism Spectrum Disorder: A Multi-Objective Reinforcement Learning Approach

Abstract

This paper presents a novel multi-objective reinforcement learning framework for Adaptive Instructional Pacing optimization in personalized educational interventions for students with autism spectrum disorder (ASD). The proposed approach addresses critical limitations of static learning systems by dynamically adjusting educational parameters based on individual cognitive profiles and real-time performance feedback. Our framework integrates cognitive-behavioral feature extraction, temporal state modeling, and fairness-aware optimization to achieve balanced improvements in learning progress and engagement task completion. An experimental evaluation on a dataset comprising 847 ASD learners across diverse cognitive presentations demonstrates significant performance gains, with learning progress improvements of 34.7% and engagement. Task completion rate improved from 72.3% to 89.3%. The algorithm maintains Statistical Parity Ratio (SPR) across different ASD presentations while achieving computational efficiency suitable for real-time deployment. Integration with Individualized Education Program (IEP) workflows demonstrates practical feasibility, with educator training requirements reduced by 42% compared to those of existing adaptive systems.