Practical and ethical challenges of large language models in education: A systematic scoping review

Abstract Educational technology innovations leveraging large language models (LLMs) have shown the potential to automate laborious process of generating and analysing textual content. While various been developed a range educational tasks (eg, question generation, feedback provision, essay grading), there are concerns regarding practicality ethicality these innovations. Such may hinder future research adoption LLMs‐based in authentic contexts. To address this, we conducted systematic scoping review 118 peer‐reviewed papers published since 2017 pinpoint current state on using LLMs support tasks. The findings revealed 53 use cases for automating education tasks, categorised into nine main categories: profiling/labelling, detection, grading, teaching support, prediction, knowledge representation, feedback, content recommendation. Additionally, also identified several practical ethical challenges, including low technological readiness, lack replicability transparency insufficient privacy beneficence considerations. were summarised three recommendations studies, updating existing with state‐of‐the‐art GPT‐3/4), embracing initiative open‐sourcing models/systems, adopting human‐centred approach throughout developmental process. As intersection AI is continuously evolving, this study can serve as an essential reference point researchers, allowing them leverage strengths, learn from limitations, uncover opportunities enabled by ChatGPT other generative models. Practitioner notes What currently known about topic Generating text‐based time‐consuming Large capable efficiently unprecedented amount completing complex natural processing generation increasingly used develop technologies that aim analysis content, such automated scoring. paper adds A comprehensive list different could potentially benefit through automation. structured assessment seven important aspects established frameworks. Three studies implement Implications practice and/or policy Updating further reduce manual effort required adapting reporting standards empirical aims need be improved. Adopting contribute resolving challenges education.