A handful of scientific publishers are investigating whether artificial intelligence can aid in peer review, Wired reports.
It writes that Isabelle Augenstein, a postdoc at University College of London, is working with Elsevier, which publishes some 7,500 journals, to develop computational tools to use in peer review, including ones for zeroing in on appropriate people to conduct that review by scanning manuscripts for key words. In addition, Wired notes that Elsevier has a suite of tools, called Evise, that checks manuscripts for possible instances of plagiarism, evaluates reviewers for conflicts of interest, and mediates workflow between editors, reviewers, and researchers.
At the same time, Springer Nature is piloting the use of a software tool called StatReviewer to assess statistical data in submitted manuscripts for completion and accuracy, Wired says.
While some are concerned about the ramifications of what Wired dubs Dr. Roboto, PhD, others note that AI has typically struggled with parsing natural language — and may have additional problems making sense of the passive voice and technical jargon found in scientific manuscripts.