Skip to Main content Skip to Navigation
Theses

Modeling the complex system of scientific publication

Abstract : The peer-review system is undoubtedly the gold standard of scientific publication. Peer review serves a two-fold purpose; to screen out of publication articles containing incorrect or irrelevant science and to improve the quality of the ones deemed suitable for publication. Moreover, the role of the scientific journals and editors is to ensure that valid scientific knowledge is disseminated to the appropriate target group of scientists and to the public. However, the peer review system has recently been criticized, in that it is unsustainable, inefficient and slows down publication. In this PhD thesis, I used complex-systems modeling to study the macroscopic behavior of the scientific publication and peer-review systems. In my first project, I modeled empirical data from various sources, such as Pubmed and Publons, to assess the sustainability of the system. I showed that the potential supply has been exceeding the demand for peer review by 15% to 249% and thus, the system is sustainable in terms of volume. However, 20% of researchers have been performing 69% to 94% of the annual reviews, which emphasizes a significant imbalance in terms of effort by the scientific community. The results provided evidence contrary to the widely-adopted, but untested belief, that the demand for peer review over-exceeds the supply, and they indicated that the majority of researchers do not contribute to the process. In my second project, I developed a large-scale agent-based model, which mimicked the behavior of the conventional peer-review system. This model was calibrated with empirical data from the biomedical domain. Using this model as a base for my third project, I developed and assessed the performance of five alternative peer-review systems by measuring peer-review efficiency, reviewer effort and scientific dissemination as compared to the conventional system. In my simulations, two alternative systems, in which scientists shared past reviews of their rejected manuscripts with the editors of the next journal to which they submitted, performed equally or sometimes better in terms of peer-review efficiency. They also each reduced the overall reviewer effort by ~63%. In terms of scientific dissemination, they decreased the median time from first submission until publication by ~47% and diffused on average 10% to 36% more scientific information (i.e., manuscript intrinsic quality x journal impact factor) than the conventional system. Finally, my agent-based model may be an approach to simulate alternative peer-review systems (or interventions), find those that are the most promising and aid decisions about which systems may be introduced into real-world trials.
Complete list of metadatas

Cited literature [163 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-02117693
Contributor : Abes Star :  Contact
Submitted on : Thursday, May 2, 2019 - 3:13:10 PM
Last modification on : Friday, March 27, 2020 - 2:43:58 AM
Document(s) archivé(s) le : Monday, September 30, 2019 - 8:07:33 PM

File

va2_Kovanis_Michail.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-02117693, version 1

Collections

Citation

Michail Kovanis. Modeling the complex system of scientific publication. Library and information sciences. Université Sorbonne Paris Cité, 2017. English. ⟨NNT : 2017USPCB034⟩. ⟨tel-02117693⟩

Share

Metrics

Record views

84

Files downloads

34