Scientific collaborator recommendation: A systematic literature review

The exponential growth of scientific literature and increasing interdisciplinary nature of research have made the manual discovery of relevant collaborators a significant challenge. Scientific collaborator recommendation systems, which apply artificial intelligence (AI) and network science to analyze scholarly data and suggest potential research partners, have emerged as an important tool to address this need. This paper presents a systematic literature review (SLR) of 70 primary studies published between 2013 and 2026 to map the evolution, synthesize the methodologies, and identify persistent gaps in this domain. These papers have been selected from a search of 1452 papers. Our analysis reveals a clear methodological progression from simple collaborative filtering (CF) to sophisticated hybrid recommendation models integrating network representation learning (NRL). We provide a critical comparative framework evaluating techniques, datasets, scholar attributes, and evaluation practices. A key finding is a pronounced utility–evaluation gap, that is, even though algorithmic performance on historical data has advanced, there is a near-total absence of user-centered evaluation, with less than 10% of studies assessing real-world utility, serendipity, or long-term impact. The review further identifies persistent challenges including dynamic network modeling, data sparsity and cold-start scenarios, and integration of heterogeneous data sources. Based on the synthesis, we contribute novel, evidence-based practical guidelines for dataset selection, attribute engineering, and system design.