Artificial intelligence (AI) is already being used in scientific publishing for a range of functions: from identifying peer reviewers and detecting plagiarism, to checking for poor reporting, data fabrication and inappropriate statistical testing. In a SpotOn Report published this month, Chadwick DeVoss, Founder and President of StatReviewer, discussed the future of automation in peer review and publishing, including other potential roles for this technology.
AI is currently being investigated for uses such as verification of author identity, prediction of article impact factor and keyword suggestion. DeVoss suggests that in the future, software may allow full automation of the publication process, including review of articles and the decision to publish. However, dehumanization of the publishing process will require vigilance. Human editors and reviewers provide an idea of what is “important” research and also ensure that research “noise” is filtered out.
DeVoss believes that automated publishing would speed up scientific communication and remove human bias. The huge volume of research that is clamouring for attention from journals means that there are currently not enough human peer reviewers to cope. AI could ease the burden on journals by finding new peer reviewers or creating automated reviews. In the future, AI may even be able to note how new research adds to knowledge in a specific field and suggest experiments to extend the research.
However, DeVoss urges caution: “The point, at which an unsupervised AI determines the direction of scientific research, is one we have to be wary of. True discovery should be an entirely human idea.”