| 1 Data collection | Scopus and WoS: 1,364 core publication articles | Search period through 2024 with search terms (“generative ai”) AND (educ*) OR (generative ai”) AND (teach*) OR (“generative ai”) AND (learn*) OR (“generative ai”) AND (student*); |
| | | Search terms were interchanged with (ChatGPT) AND (educ*) OR (ChatGPT) AND (teach*) OR (ChatGPT) AND (learn*) OR (ChatGPT) AND (student*) |
| 2 Quality checks | Scopus and Web of Science | To guarantee that documents are related to generative AI in Education, all abstracts were read: 951 final documents |
| 3 SMS analysis and cluster identification | VOSviewer and RStudio | Network map of 817 documents based on keywords co-occurrence; four core clusters were automatically identified. The applied techniques enable the illumination of large research literature (Waltman et al., 2010) and have already been used in educational papers (e.g. Kosmützky and Krücken, 2014; Tseng et al., 2013) |
| 4 Further analysis and maps | Scopus and WoS, VOSviewer and RStudio | Co-occurrence of keywords and Scopus and WoS research areas; development of the generative AI in education research literature. These additional maps provide further detail to the taxonomy scheme and a cross-check of the results |
| 5 Cluster interpretation | Content analysis and qualitative interpretations | A content analysis was conducted to name and define the four clusters of generative AI in education. This qualitative interpretation involved reviewing 817 articles, focusing on their titles, authorship and data sources to identify distinct research streams, in line with methodologies proposed by prior studies (e.g. Gaur and Kumar, 2018) |
| 6 Taxonomy scheme | | The analyses were grouped into a taxonomy scheme, with each cluster analyzed and discussed to identify future research recommendations |