Correlation between references and citations in artificial intelligence: A preliminary study

Abstract

To reveal whether there exists a correlation between the number of references and the number of citations in a paper of the Artificial Intelligence‐related conference, we present an experiment based on a total number of 9,458 papers published in CVPR, ICML, and AAAI from the year 2014 to 2018. Spearman's rank correlation tests show that the number of references and that of the citations are correlated with a correlation coefficient of 0.435, 0.464, and 0.180 respectively, in each p < .01. This means an increasing number of references in a paper may be contributing to the increase in citations. However, citing more references does not guarantee that a paper will be a highly cited paper. We hope this finding can provide a basis for the prediction of the citation impact of the Artificial Intelligence conference papers in the future.

Factors associating with or predicting more cited or higher quality journal articles: An Annual Review of Information Science and Technology (ARIST) paper

“…Articles with more references are more cited for Biology and Biochemistry, Chemistry, Mathematics, Physics (Vieira & Gomes, 2010), ecology (Mammola et al, 2021), clinical articles from medical journals (Lokker et al, 2008), AI (Xiao & Jiang, 2020), psychology (Haslam et al, 2008), psychiatry (Hafeez et al, 2019), library and information science (Yu et al, 2014), management (Antonakis et al, 2014), tourism, leisure and hospitality (Cunil et al, 2023), and six biomedical topics (Urlings et al, 2021). This seems to be a universal pattern, perhaps because a longer reference list suggests connections to a wider literature (and therefore potentially more widely relevant), higher quality research (because more justified through references) or a longer article.…”