Artificial intelligence in classroom discourse: A systematic review of the past decade

Over the past decade, multiple artificial intelligence-based models and systems have been developed and employed in classroom discourse as a means of facilitating learning and teaching. To provide valuable guidance for future studies seeking to effectively integrate powerful artificial intelligence (AI) technologies, this paper presents a systematic review of the literature on artificial intelligence in classroom discourse (AICD) over the past decade. Following the latest PRISMA framework, we searched the Web of Science database and the relevant conference proceedings and identified a total of 68 studies. Five themes across the studies were examined: basic sample statistics, educational contexts, data sources, AI technologies, and the effects of AICD. The findings revealed that most AICD studies focused on science-related and language-related at the primary and secondary school levels. Various AI models and systems were developed and used to analyze student-related interaction and learning (e.g., speech acts and collaboration), teacher-related instructional behavior (e.g., question-asking and uptake), and whole-class dialogue (e.g., topic evolution) based on their discourse data in both online and offline classroom settings. Furthermore, the use of AI in classroom discourse was found to be impactful on learning gains, emotions (e.g., willingness and affordance), behavior (e.g., teachers' uptake and students' conversation), and perception of AI. We further discussed the future directions of the field, such as the increasing popularity of deep learning models (e.g., ChatGPT) in investigations of classroom discourse and the urgent need to evaluate the effect of AICD in practice.