[Paper Review] Copied citations create renowned papers?

M. V. Simkin, Vwani Roychowdhury|arXiv (Cornell University)|May 8, 2003

scientometrics and bibliometrics research21 citations

TL;DR

This paper proposes that the citation distribution in scientific literature—where a few papers receive vastly more citations than others—can be explained by a simple probabilistic model based on citation copying, not scientific merit. The model shows that when researchers randomly select papers and copy references from them (including 25% of cited references), the resulting citation patterns match empirical data, suggesting that popularity arises from stochastic copying, not genius.

ABSTRACT

Recently we discovered (cond-mat/0212043) that the majority of scientific citations are copied from the lists of references used in other papers. Here we show that a model, in which a scientist picks three random papers, cites them,and also copies a quarter of their references accounts quantitatively for empirically observed citation distribution. Simple mathematical probability, not genius, can explain why some papers are cited a lot more than the other.

Motivation & Objective

• To investigate whether citation copying, rather than scientific quality, explains the heavy-tailed distribution of citations in scientific literature.
• To model the mechanism by which scientists propagate citations by copying reference lists from other papers.
• To test whether a stochastic model based on random selection and copying can reproduce empirically observed citation distributions.
• To determine the extent to which citation popularity is driven by network effects and copying behavior rather than original impact.
• To challenge the assumption that highly cited papers are cited because they are inherently more important or innovative.

Proposed method

• Propose a stochastic model in which a scientist selects three random papers and cites them.
• Incorporate a 25% probability of copying references from each cited paper into the new reference list.
• Simulate citation propagation across a network of papers using this copying mechanism.
• Compare the resulting citation distribution from the model to real-world citation data from scientific literature.
• Use mathematical probability theory to derive the expected citation frequency distribution under the copying model.
• Validate the model by fitting it to empirical citation data and assessing goodness of fit.

Experimental results

Research questions

• RQ1 To what extent does citation copying explain the heavy-tailed distribution of citations in scientific literature?
• RQ2 Can a simple probabilistic model based on random selection and copying replicate observed citation patterns?
• RQ3 What fraction of references must be copied to reproduce real citation distributions?
• RQ4 Does the model explain why a small number of papers receive a disproportionately large number of citations?
• RQ5 Is the observed citation popularity in science primarily driven by stochastic copying rather than intrinsic scientific value?

Key findings

• The model, which includes copying 25% of references from cited papers, quantitatively reproduces the empirically observed citation distribution in scientific literature.
• The citation distribution generated by the copying model closely matches real-world data, indicating that stochastic copying alone can explain citation inequality.
• The model demonstrates that high citation counts can emerge purely from random selection and copying, without requiring scientific excellence.
• The results suggest that a significant portion of citation impact is due to network effects and reference list propagation, not original contribution.
• The model's success implies that citation counts may be poor indicators of scientific quality or originality.
• The study challenges the assumption that highly cited papers are cited because they are more important, showing instead that copying behavior can create 'renowned' papers through chance.