Characterizing and modeling citation dynamics (1110.2153v1)

Abstract: Citation distributions are crucial for the analysis and modeling of the activity of scientists. We investigated bibliometric data of papers published in journals of the American Physical Society, searching for the type of function which best describes the observed citation distributions. We used the goodness of fit with Kolmogorov-Smirnov statistics for three classes of functions: log-normal, simple power law and shifted power law. The shifted power law turns out to be the most reliable hypothesis for all citation networks we derived, which correspond to different time spans. We find that citation dynamics is characterized by bursts, usually occurring within a few years since publication of a paper, and the burst size spans several orders of magnitude. We also investigated the microscopic mechanisms for the evolution of citation networks, by proposing a linear preferential attachment with time dependent initial attractiveness. The model successfully reproduces the empirical citation distributions and accounts for the presence of citation bursts as well.

• The paper presents a refined model using KS statistics that shows the shifted power law best describes citation distributions in APS journals.
• It identifies citation bursts as early rapid accumulation periods that deviate from traditional linear preferential attachment.
• The study proposes a time-dependent attractiveness model, challenging existing metrics and highlighting temporal variations in research impact.

Analysis and Modeling of Citation Dynamics

In "Characterizing and Modeling Citation Dynamics," Young-Ho Eom and Santo Fortunato present an in-depth analysis of citation distributions, with a focus on papers published by the American Physical Society (APS). Their research endeavors to identify the functional form that best describes citation distributions over various time spans, facilitating a better understanding of the underlying mechanisms of citation dynamics.

The authors compare three hypothesized models: log-normal, simple power law, and shifted power law, assessing their goodness of fit using Kolmogorov-Smirnov (KS) statistics. They conclude that the shifted power law provides the most reliable description across different time windows. This finding is pivotal, revealing that while previous studies proposed different distribution models, the shifted power law adequately captures the complexity of citation dynamics in APS journals, specifically noting that the power-law exponents decrease over time.

A distinctive feature of citation dynamics identified in this work is the presence of "bursts," instances of rapid citation accumulation occurring predominantly within the initial years following publication. These bursts demonstrate a non-uniform attraction pattern that deviates from the steady patterns predicted by traditional models like linear preferential attachment. To address this discrepancy, Eom and Fortunato propose a refined model that incorporates a time-dependent initial attractiveness in addition to linear preferential attachment. This model aligns well with their empirical data, supporting both the observed distributions of citation metrics and the presence of citation bursts.

The implications of this paper extend to several domains within bibliometrics and network science. The elucidation of shifted power law distributions as a dominant pattern challenges prior assumptions about citation dynamics and suggests that citation strategies and metrics may require reconsideration, particularly those relying on stable, predictable models. Furthermore, the emergence of citation bursts underscores the necessity of adopting nuanced models that account for temporal variations in research impact and reception.

Looking towards the future, this work invites further exploration into the variables influencing initial attractiveness and its decay over time. Such research could provide deeper insights into factors contributing to a paper's 'burstiness,' ranging from social to scientific elements germane to a given field. As citation networks continue to evolve with increasing complexity, integrating findings like these with advancements in artificial intelligence and machine learning could yield new predictive models, enhancing our understanding of scientific influence and knowledge dissemination.

Overall, Eom and Fortunato's research stands as a significant contribution to the paper of citation networks, offering a more comprehensive framework for analyzing the intricate patterns of scientific literature dissemination and impact across time.