Adaptive AI for Science Text Accessibility in Students with Reading Disabilities

• In Event: Bridging Gaps: AI and Digital Learning for Diverse Learners

Abstract

This study explores the use of neuroadaptive artificial intelligence (AI) to support students with dyslexia in reading science content. Using functional near-infrared spectroscopy (fNIRS) and a convolutional neural network (CNN), we classified cognitive demand in real time and dynamically adjusted science text features. Results from 100 students (50 with dyslexia) showed significant gains in reading comprehension and test performance when text was adapted based on neurocognitive feedback. The CNN achieved 86% classification accuracy, enabling personalized support. Findings highlight the potential of neuroadaptive AI to enhance accessibility and learning outcomes for students with reading disabilities, with implications for inclusive science education and future AI design.

Authors

• Richard Lamb, University of Georgia

• Danielle Jo Malone, Purdue University

• Tosha L. Owens, East Carolina University

• Ikseon Choi, Emory University

• Shruti Kundu, University of Georgia