User control of search algorithms would improve science

Abstract

When scientists search for articles using platforms such as Google Scholar, Scopus and Web of Science, the results that they see are determined by search algorithms. These algorithms prioritize highly cited papers. On the one hand, this strengthens the association of these papers with corresponding keywords, which helps to surface dominant publications from prevailing disciplines. However, it also narrows exposure to alternative methods, theories or research programmes. Consequently, researchers who explore emerging topics or interdisciplinary problems face diminished chances of discovering relevant but less-cited perspectives.

These scientific ‘recommender systems’ shape our information environments, and amplify popular items through self-reinforcing feedback loops¹. Items that initially gain prominence receive disproportionately greater visibility, and gain popularity simply because they are already popular — regardless of their quality or utility². This inadvertently constrains individuals’ potential to engage with a fuller spectrum of ideas.