Manual versus AI-assisted document screening (ASReview): a comparative analysis within a rapid systematized review in the social sciences

This article offers a nuanced evaluation, in the context of a rapid systematized review in the social sciences, of manual screening and artificial intelligence (AI)-assisted screening using ASReview. Drawing on a corpus of 2,800 references on strategic planning for senior living environments, a mixed quantitative and qualitative approach was used. The results indicate an increase in agreement between methods, with the Jaccard coefficient rising from 0.407 after initial screening to 0.710 following re-evaluation. AI-assisted screening increased efficiency by reducing the time required (43 hours vs. 9 hours), while revealing qualitative differences in the profiles of the articles selected: manual screening favors contextual and qualitative studies, while AI-assisted screening captures more quantitative or international studies. These exploratory findings highlight the complementarity of human expertise and algorithmic support, and suggest that hybrid strategies may best reconcile rigor, diversity, and efficiency in systematized reviews in the social sciences.