The effectiveness of ‘peer review’ — the system whereby scientists critique each other’s research to determine whether or not it should be published — is investigated in new findings published in Nature.
Researchers from the Universities of Bath and Bristol suggest that the system, which is one of the main ways in which science attempts to ensure quality research, is open to flaws but identify an improved model that would better ensure the quality and veracity of published work.
Using a mathematical model of the behaviour of scientists, the team aimed to better understand how the model of peer review works, and in particular whether it may be susceptible to failure by examining the factors that can influence scientists when critiquing the work of others.
As reviewers are often encouraged to consider only largely objective characteristics of the study they are reviewing, the team found that they are open to "herding" — a phenomenon whereby scientists’ behaviour may be influenced, and even dominated, by information gleaned for their peers’ behaviour rather than by their own personal opinions. The researchers found that this increases the risk of reviewers converging on what they perceive to be a correct answer but is actually an incorrect answer.
The study found that peer review performs best when a degree of subjectivity (i.e., their belief about whether the result is correct) is allowed, since this enables more information to be transmitted through the decision and protects against the risk of scientists converging on an incorrect answer.
An improved peer review model includes the opportunity for scientists to comment on and critique research after it has been published. This would provide scientists with more opportunity to truthfully reveal their opinions, and improve the flow of information in science and avoid the risk of herding.
Mike Peacey, one of the study’s researchers who is based jointly at the Universities of Bath and Bristol, said: “Scientists are increasingly concerned that many published research findings may be false. Our findings demonstrate a novel way of how the peer review process can be improved through a post-publication peer review.”
The study was funded by the Economic and Social Research Council and the Medical Research Council.