Maciej J. Mrowiński, Piotr Fronczak, Agata Fronczak, Marcel Ausloos, Olgica Nedic

Peer review is one of the cornerstones of scientific publishing. It is a process whose goal is to ensure (and often improve) the quality of scientific work published in journals. However, despite its importance, peer review remains an understudied subject and only in recent years has it attracted the attention of scientists interested in the dynamical aspect of the process.

On of the problems in peer review is that the number of manuscripts submitted for publication is increasing every year. This puts strain on reviewers and leads to longer review time.

In our work, we use Cartesian Genetic Programming [1] to improve the effectiveness (that is to reduce review time) of peer review. Cartesian Genetic Programming is a technique which employs (much like genetic algorithms) certain concept known from the theory of evolution in order to artificially evolve optimised solutions to user-defined problems. We use CGP to evolve editorial strategies – sets of rules that help editors decide when (and how many) invitations should be sent to potential reviewers.

Our approach is entirely data-driven: we use the analysis presented in [2] as the basis for simulations of the peer review process. These simulations allow the CGP algorithm to assess and optimise editorial strategies. The strategies we managed to evolve result in review time shorter even by 30% when compared to typical strategies used by actual editors. We also show that by employing groups of mixed reviewers – that is reviewers know personally by the editor and reviewers found through scientific databases – review time can be shortened even further.

[1] Miller JF. Cartesian Genetic Programming. Springer, 2011.

[2] Mrowinski MJ, Fronczak A, Fronczak P, Nedic O, Ausloos M. Review time in peer review: quantitative analysis and modelling of editorial workflows. Scientometrics. 2016, 107(1):271-286.