Wikipedia and open access

Puyu Yang, Ahad Shoaib, Robert West, Giovanni Colavizza

TL;DR

This paper investigates how open access (OA) influences the likelihood of scientific articles being cited on Wikipedia. It finds that OA articles are significantly more likely to be cited, with this effect being stronger for newer, highly cited articles. This suggests OA plays a crucial role in disseminating scientific knowledge to a broader audience through platforms like Wikipedia.

What Problem Does This Paper Address?

Wikipedia is a massive repository of information, and its reliability hinges on the quality of its sources, particularly academic publications. The open access movement aims to make scientific research freely available, which intuitively should benefit Wikipedia. However, the extent to which OA status actually impacts an article's visibility and citation on Wikipedia, especially at the individual article level, has not been thoroughly explored. Previous studies often focused on journals rather than individual articles, potentially underestimating the impact of OA, particularly for articles available through "green" or "hybrid" open access routes. This paper aims to fill this gap by analyzing Wikipedia citations at the article level to understand the relationship between OA and citation patterns.

Methods

The researchers employed a multi-stage data collection and analysis process. They started with a large dataset of citations from English Wikipedia. To enrich this data, they utilized the OpenAlex API to gather article-level information such as OA status, publication date, and citation counts. Journal-level data was obtained from Scimago.

The core of their analysis involved statistical modeling, specifically logistic regression. They developed two main models to assess the influence of OA on an article's likelihood of being cited in Wikipedia. The first model focused on the direct effect of OA status, while the second model incorporated citation counts to explore mediated effects. They also performed a multicollinearity check to ensure the robustness of their models. To address potential biases and ensure a balanced dataset for regression, they used stratified sampling to create negative samples (articles not cited in Wikipedia).

Datasets

  • Wikipedia Citations Dataset: A comprehensive dataset of over 45 million citations extracted from the English Wikipedia February 2024 dump. Approximately 2.2 million of these are classified as journal articles.
  • OpenAlex API: Used to retrieve article-level data, including OA status, publication date, publisher, and concepts, for journal articles identified by DOIs.
  • Scimago: Used to acquire journal-level information, such as SJR (Scimago Journal Rank) score and H-index.

The study constructed a balanced dataset for regression analysis, combining Wikipedia-cited articles with carefully sampled negative samples.

Key Insights and Ideas

  • OA as a Wikipedia Citation Driver: The study strongly suggests that open access significantly increases the probability of a scientific article being cited on Wikipedia. This is a direct benefit of the OA movement for knowledge dissemination.
  • Beyond Journal-Level Analysis: By focusing on individual articles, the research provides a more nuanced understanding of OA's impact, capturing articles available through various OA routes that might be missed in journal-level analyses.
  • Newer, Highly Cited Articles Benefit Most: The OA citation advantage is particularly pronounced for articles that are more recent and have higher citation counts, indicating that influential and up-to-date research is more likely to be discovered and utilized on Wikipedia.
  • Wikipedia as a Science Communicator: The findings highlight Wikipedia's role in amplifying the reach of scientific knowledge, making it accessible to a wider audience, especially through OA content.

Key Results

  • 64.7% Higher Likelihood of Citation: Open access articles have a 64.7% higher likelihood of being cited on Wikipedia compared to closed-access articles, after controlling for confounding factors.
  • OA Prevalence: 46.5% of all scientific citations on Wikipedia are from OA articles, a notable increase from previous estimates.
  • Growing OA Trend: The proportion of OA articles cited by Wikipedia has been consistently rising over the past four decades, surpassing 50% after 2015.
  • Concept-Specific Variations: While OA articles are generally favored, there are variations across scientific concepts. Biology, Physics, and Mathematics show higher OA citation rates, while Social Sciences and Humanities have lower rates.
  • Article Age and Citation Count Matter: Newer articles (published within the last ~4 years) and articles with higher citation counts are significantly more likely to be cited on Wikipedia, with OA status further amplifying this effect.

Limitations and Future Work

  • Exclusion of Certain Sources: The study focused on articles with DOIs, which excluded conference papers and earlier literature. Future research could incorporate these sources.
  • Time Dimension: The analysis did not explicitly incorporate time as an analytical dimension. Examining Wikipedia's edit history for specific citation timestamps could provide deeper insights into the causal mechanisms.
  • Other Causal Variables: While significant factors were controlled for, other variables like article length might influence citation patterns and could be explored in future work.
  • Broader Open Science: The authors suggest extending this research to other forms of open science, such as open research data and software, to understand their impact on platforms like Wikipedia.

Summary updated

BibTeX
@article{yang2024wikipedia,
  author = {Puyu Yang and Ahad Shoaib and Robert West and Giovanni Colavizza},
  title = {{Wikipedia and open access}},
  journal = {Scientometrics},
  year = {2024},
  doi = {10.1007/s11192-024-05163-4},
}
Figure