Generative artificial intelligence in special education: A systematic review through the lens of the mediated-action model
Meiqin Wang, Ahmed Tlili, Mohamed Koutheaïr Khribi, Chung Kwan Lo, Ronghuai Huang
Abstract
Generative Artificial Intelligence (GenAI) has gained increasing attention from researchers and practitioners in various domains including education. However, little is known about how GenAI is being used in special education as well as the effects and challenges that might bring for learners with disabilities. To address this research gap, this study systematically reviews the literature on GenAI in special education. Specifically, it conducts a meta-synthesis of thirty-three studies based on Vygotsky's mediated-action model. The obtained results revealed that: (a) most of the studies on GenAI in special education are from the U.S. and the most targeted disability type is emotion/attention; (b) GenAI was mostly used as a pedagogical assistant; (c) although a quantitative synthesis on experimental studies indicated a huge effect ( g = 1.49) of GenAI on learning outcomes, the statistical significance remains inconclusive, suggesting the need for further research in this area; and (d) despite the opportunities brought by GenAI in special education, many concerns are found, including ethical, competency, and technical. The findings of this study can support various stakeholders (e.g., policy makers, educators, etc.) adopt safely and effectively GenAI in special education, hence contributing to a more accessible and inclusive education.