Honglin Bao and Misha Teplistkiy (Michigan). 

Preprint (Under review); Twitter thread; Slides

TL; DR: The scientific community generally discourages authors of research papers from citing papers that did not influence them because such "rhetorical" citations are assumed to degrade the literature and incentives for good work. Intuitively, a world where authors cite only substantively appears attractive. We evaluate this intuition by utilizing agent-based models to construct a counterfactual world with substantive citing only and we show that, surprisingly, rhetorical citing benefits academic community health, deconcentrates attention, and makes it easier to displace incumbent ideas. The proximate explanation for the effect is that the quality of creative products, like papers, is hard to discern and people thus use heuristics to judge them. In a world with substantive citing only, citations and attention would be concentrated among the highest-status papers (quality, citation, etc.), and that concentration would increase via the feedback loop of researchers choosing highly cited works to read and citing them yet more (like the Matthew effect). However, the practice of rhetorical citing breaks this reinforcing loop by redistributing citations from the few elite-quality papers to a more diverse set that is rhetorically useful in persuading readers or reviewers (like supporting the citers' own claims) in a person-specific manner.