Peer-review is an important part of scientific communication. It is not a panacea and is quite fallible, but it serves several functions well:
- It gives journals a “filter” against unprofessional work (including crankery and its cousins)
- It gives authors and editors the opportunity to receive unfiltered feedback on how a paper will be received. The “unfiltered” part is ensured in part by the anonymity of the referee (which the referee can waive).
- It provides a fresh set of eyes from an expert who can help spot errors and mistakes
- Ideally, it gives authors constructive feedback to improve their work
- It does all of this confidentially, so when science is finally presented to the world it has already been vetted and mistakes have been fixed (this step is optional for authors—they can post preprints whenever they like).
Refereeing papers is an unpaid service to the community that scientists provide. I think of it like jury duty: I rely on good referees to fairly judge and improve my papers, and so I try to write at least as many referee reports as I receive, and be the sort of referee I hope I’ll get on my papers (I have to referee more than I am refereed because some scientists are very bad referees (slow, delinquent, unduly harsh, sloppy, etc.) and the rest of us need to pick up their slack.
The confidentiality of the process is important to the journals, and to some authors and referees. For the authors, this means they do not need to worry that their work will be scooped, that their mistakes have a chance to be fixed before they are made public, and that they have a chance to respond to any unfair criticisms in a report. For journals, this confidential review process is one of the primary values they add to papers (some of the others being editing, typesetting, publicity, and curation). For referees, it means they can frankly help improve the literature without being concerned about the reaction of the paper’s authors.
There are other ways to do this; there are good arguments for entirely open peer review, as well.
One interesting issue is whether the referee should be bound by confidentiality. There are good reasons why a referee might feel they should publish their review. This is especially true in instances where there is evidence of fraud, for instance. The case is made here:
"As a Reviewer, Each Review is Yours, and Sometimes It Should Be Everyone’s", by @lteytelman https://t.co/r2L2eID6Fh
— Peter Edmonds (@PeterDEdmonds) April 2, 2017
The argument is that there can be good reasons to publish your report (as open peer review shows; read the article), and because the referee holds the copyright to the review.
But in the AAS astronomy journals, confidentiality of referees report is explicitly part of the journal ethics statement that binds referees and authors. Referees thus promise to keep the reports confidential.
Of course, this could lead to an ethical dilemma if publishing the referees report would improve science. But I think there are many perfectly ethical routes one can take to avoid that dilemma:
- Polish the relevant parts of your referee’s report, make sure it only refers to the published paper, and not the (strictly confidential) manuscript, and publish that on the arXiv as a “comment”.
- If the content of the manuscript needs to be mentioned, go directly to the editor. The purpose of confidentiality is to preserve scientific integrity; the editor has the authority to waive confidentiality, or take other actions to rectify the problem.
For instance, a climate science journal Environmental Science Letters rejected a paper by a climate skeptic, Lennart Bengtsson, on the basis of two very unfavorable referees reports that heavily criticized the science. The reports even offered suggestions for how to improve the paper and make it publishable.
Prof. Bengtsson cherry-picked a few words from one of the reports and (violating confidentiality) complained to the media that the paper had been rejected because of its conclusions for political reasons.
The journal, in this case published a rebuttal that included the entire referees reports, exposing the mendacity of Prof. Bengtsson’s accusations. Review confidentiality is at the discretion of the editors, so they can (and should) waive it when it is not serving its purpose (preserving scientific integrity).
If, for some reason, the editors were to enforce the confidentiality to the detriment of scientific integrity, then a referee may need to violate the rules and publish their review. In this case, I recommend checking (confidentially) with others familiar with the issues, especially an ethicist, to make sure you are in the right. After you post your report, the journals may complain, and you may face repercussions from your professional society, but if on balance you are behaving ethically you will have a good defense.