From joy to fear in scientific titles: automated emotion recognition and the citation payoff

Citation-based indicators are widely used to assess scholarly impact, yet they may be linked to factors unrelated to scientific contribution. This study investigates whether emotions in article titles might constitute an attentional bias that could be associated with citation outcomes. Focusing on marketing research, we analyze 18,272 article titles indexed in Scopus, of which 15,836 are included in the final analysis after data cleaning. We assess their emotional content with ChatGPT by aggregating the outcomes of multiple runs rather than depending on a single output, and we organize the resulting classifications according to Plutchik’s psychoevolutionary model of emotions. Using large language models helps us capture subtle differences in emotional intensity and avoid the limits of traditional lexicon-based methods. We estimate Pseudo Poisson Maximum Likelihood models with journal and publication-year fixed effects to assess associations between emotional signals and citation counts. The results show that certain primary emotions, such as joy, as well as specific emotional combinations, are systematically associated with higher citation performance, while others, like despair, correlate with lower visibility. These results indicate that the emotional complexity of scientific titles is associated with citation patterns, pointing to an affective factor that has received little attention in how scholarly visibility and impact are constructed.

Affiliations

UCLouvainSSH/LouRIM - Louvain Research Institute in Management and Organizations

Citations

APA
Chicago
FWB

Gerardy, N., Kervyn de Meerendré, N., & Verardi, V. (2026). From joy to fear in scientific titles: automated emotion recognition and the citation payoff. Scientometrics. Published. https://doi.org/10.1007/s11192-026-05720-z (Original work published 2026)