Title: Fearing the machine: An evaluation of the sensitivity and specificity of machine learning tools for abstract screening in systematic reviews in education
Abstract: Meta-analyses and systematic reviews are essential for developing practice, science, and policy. However, publishing a meta-analysis could require more than a year of work. Therefore, current systematic reviews and meta-analytic approaches are not sustainable because the review of recent evidence cannot be made efficiently and quickly before becoming obsolete. Artificial intelligence (AI) and machine learning algorithms can support researchers in speeding up the literature search process by identifying and screening relevant records for their systematic review. However, the performance of those tools in the educational sciences is unknown. In this study, we aim to provide researchers with information on the sensitivity and specificity of AI screening tools within the educational domain. We hope this study provides answers for applied researchers about the performance of AI tools for literature screening.