A Systematic Literature Review of Empirical Research on Applying Generative Artificial Intelligence in Education
Xin Zhang,Peng Zhang,Yuan Shen,Min Liu,Qiong Wang,Dragan Gašević,Yizhou Fan
DOI: https://doi.org/10.1007/s44366-024-0028-5
2024-01-01
Abstract:Generative artificial intelligence (GenAI), achieving human-like capabilities in interpreting, summarising, creating, and predicting language, has sparked significant interest, leading to extensive exploration and discussion in educational applications. However, the frontline practice of education stakeholders or the conceptual discussion of theorists alone is not sufficient to deeply understand and reshape the application of GenAI in education, and rigorous empirical research and data-based evidence are essential. In the past two years, a large number of empirical studies on GenAI in education have emerged, but there is still a lack of systematic reviews to summarise and analyse the current empirical studies in this field to evaluate existing progress and inform future research. Therefore, this work systematically reviews and analyses 48 recent empirical studies on GenAI in education, detailing their general characteristics and empirical findings regarding promises and concerns, while also outlining current needs and future directions. Our findings highlight GenAI’s role as an assistant and facilitator in learning support, a subject expert and instructional designer in teaching support, and its contributions to diverse feedback methods and emerging assessment opportunities. The empirical studies also raise concerns such as the impact of GenAI imperfections on feedback quality, ethical dilemmas in complex task applications, and mismatches between artificial intelligence (AI)-enabled teaching and user competencies. Our review also summarises and elaborates on essential areas such as AI literacy and integration, the impact of GenAI on the efficiency of educational processes, collaborative dynamics between AI and teachers, the importance of addressing students’ metacognition with GenAI, and the potential for transformative assessments. These insights provide valuable guidelines for future empirical research on GenAI in education.