Key points are not available for this paper at this time.
Abstract Multimodal generative artificial intelligence is revolutionizing critical domains, including education, healthcare, business, and entertainment. This article presents a comprehensive umbrella review of prominent surveys, reviews, and SLRs on advances in generative AI in teaching and learning, synthesizing its applications, benefits, challenges, and implications. We applied the systematic literature review (SLR) protocol and PRISMA framework to guide our search methodology and screen the relevant studies to ensure accuracy and replicability and reduce bias. We searched three popular academic databases, namely Scopus, Web of Science, and Google Scholar, screening 451 candidate articles published between 2020 and 2024. From these, we identified 46 relevant GenAI reviews in education. Our investigation of the reviews answered five research questions covering the academic disciplines adopting generative AI, applications of large multimodal models in education, their impact on student performance, their impact on teaching and assessment, and overarching challenges. Recent trends showed China and the USA as leading publishers of GenAI reviews, ChatGPT as the most popular GenAI tool used in education, and text as the primary modality of interaction among students and teachers. Several reviews explored the application of generative AI in multiple academic disciplines instead of a single domain. Students use GenAI tools to understand complex notions, learn foreign languages, improve writing skills, prepare for assessments, and enhance engagement. However, teachers apply GenAI tools to develop course materials, improve pedagogical methods, design various assessments, evaluate learning outcomes, and provide personalized feedback. We conclude our umbrella review by presenting a taxonomy of challenges impeding the adoption of multimodal generative AI in teaching and learning. Key issues of GenAI include academic dishonesty, negative impacts on critical thinking skills, integration with traditional teaching philosophies, and lack of emotional intelligence in generative AI platforms.
Namoun et al. (Wed,) studied this question.