Full Library

In this paper, we present the first study that undertakes a comparison of human and Artificial Intelligence (AI) tools in a large-scale literature screening process as part of a large-scale meta-analysis in education research. We present this screening approach to AI-powered Literature Reviews (AILRs) at a large-scale, and then apply this approach to a screening of a public evidence library, containing several-hundred thousands of articles on research pertaining to SDG4 in low-income countries, of which we categorized along various dimensions with the goal of making these categorized datasets available to the wider research community. Our approach is transparent, open-source, and replicable. Importantly, it is a ‘living’ approach, enabling the ability to make incremental updates as work progresses. We believe that this project presents an important opportunity towards further developing shared knowledge on AILRs in screening processes. Furthermore, these findings hold practical significance for the broader goal of building and sharing a comprehensive, digital, and open-source evidence library on ‘what works’ in education across low-income countries. Ultimately, sharing key learnings from our research and the resulting evidence library is beneficial to both research practices and enabling collaborations in the wider research community, especially as it relates to SDG research and development projects in education.

DOI: 10.53832/opendeved.1132 ZenodoArchiveID: 11105029 ZenodoArchiveConcept: 11105028