Is artificial intelligence the key to speedier peer review?
In an age of rapid sharing of online information, authors are becoming increasingly frustrated with the lengthy timeframes associated with scientific peer review. But what if artificial intelligence (AI) could help editors to streamline their approach and speed up the process? A recent study by Mrowinski et al, published in PLOS One, highlighted the potential of Cartesian Genetic Programming (CGP) – an evolutionary algorithm inspired by natural selection and mutation – to cut peer review times by 30%.
It’s easy to see why current review times are lengthy. First an editor must identify suitable reviewers, then send invitations out. After an agreed period of time an enquiry is sent to any non-responders – over 70% of invitations receive no response within 7 days. If a reviewer responds to the invitation or enquiry and agrees to review, the editor must wait a further 25 days before contacting them for their report. At this stage, some reviewers do not respond and some choose not to review the article.
In the hope of addressing some of these issues, Mrowinski et al compared a strategy developed using CGP with two ‘traditional’ editorial strategies. ‘Traditional’ Strategy 1 involved sending out a batch of invitations and waiting for the whole process to finish for all potential reviewers in the batch before sending out another batch of invitations if necessary. ‘Traditional’ Strategy 2 involved sending out a batch of invitations, then sending out another invitation every time one reviewer completed the process, until enough reviews were received. In the CGP-based strategy, the algorithm used a number of factors to inform and adapt the method used to individual scenarios, including the number of reviewers in the batch, the number of reviewers ‘active’ for the article, the number of reviews required per article, and the number of reviews received thus far. The CGP-conceived strategy decreased review time by 17 days compared with Strategy 2, and by 6 days compared with Strategy 1. The authors explain that although the CGP strategy used worked for the specific journal studied, it is not universal. However, the method could be used to develop similar editorial strategies for other journals.
The authors also note that although AI can reduce the duration of the peer review process, the speed at which potential reviewers respond to invitations – even to decline them – has an enormous impact. So, to all scientists clamouring for speedier peer review – the power is in your hands!
Summary by Philippa Flemming, PhD from Aspire Scientific
Leave a Reply