How to fix scholarly peer review

All academics invariably suffer from an occasional bad and unfair review. We pour our sweat (and sometimes tears) over a paper for many months to watch it summarily and unfairly shot-down by a reviewer. Opening a decision email to read such a review feels so sickening that it hurts in the stomach. Even after 20 years, I am unable to desensitize myself to the pain of being shot-down by an unfair review. I suspect quite many people quit academia being frustrated over bad reviews.

Peer review is a complex and sensitive topic, and this post will inevitably fall short of capturing some important aspects of it. But I believe this topic is so important that it deserves more attention and conversation. Here I first write a bit about how to handle bad reviews. Then I outline some problems with our current peer review process and suggest some fixes to start a conversation on this.

The first rule of review club

The first rule of review club is to always give the benefit of doubt to the reviewers. We are all partial about our work: if we didn't think highly of it, we wouldn't have worked on it so hard, and submitted it to a  good venue. I suggest that you avoid responding to the reviews the first day. Chances are that you will be too busy processing your emotions to have bandwidth to process the reviews calmly. So sleep on it, read the reviews again the next day. Give it your best shot to see what gets criticized. Try to see what got misunderstood. If some reviewers had this problem, several of your readers will have similar problems as well. Can you improve your presentation to avoid that problem? How can you address/disarm the criticisms from the reviewers? What is the opportunity to improve here? The idea is to learn whatever you can learn from the reviews, and become so good they can't ignore you.

The second rule of review club

The second rule of the review club is that occasionally you get a review so malicious, so bad that you are stumped. This is inevitable. Welcome to the club!

What do you do now?

Again exercise caution. Give it a day, meet with your coauthors. Try to give the benefit of the doubt to the reviewers. If you still cannot justify in any way that the reviewer is not malicious and unfair, you should contact the PC chairs or Editor to complain about the behavior. This often gets ignored, so don't expect much.

Basically you swallow this abuse of power and move on. You try not to get too upset by many months wasted waiting for this rejection, and try to plan ahead for another publication venue.  Maybe you make a mental note to avoid this conference and journal. But isn't that penalizing yourself, and not the guilty party?

If you like to fight the unfairness, you don't have much leverage. You can make the reviews public to expose the unfairness to community. But that is about it, and this probably won't make much difference either. Yes, very frustrating. Very unfair. I know.

The current peer review process

This is how the current scholarly peer review works. You submit your paper, and it gets reviewed by (often) 3 anonymous reviewers. After 3 months (or 6 months for the journals), you get the reviews and the decision in the email. The rating is generally average of the reviewers, and a bad review damages your chances especially severely for conferences.

The blind reviewing works well when the reviewers are conscientious, and the PC-chairs/editors take meticulous care in their job to oversee the reviewers. But note that there is an asymmetric power relation here. The problem is that anonymous reviewers don't have much skin in the game, and the process is open to abuse.

Good intentions don't work, good systems work. And there is obviously something broken with our reviewing system today. The quality of reviews within the same conference or journal is uneven and varies wildly. Reviewers, who are overworked academics, are overwhelmed with lots of papers to review. Competent reviewers are especially overwhelmed. As an associate editor at IEEE TCC, when I try to find reviewers for papers, my usual success rate is 3 over 8-10. I need to contact up to 10 reviewers to find 3 that accepts to review the paper.

The basic recipe 

Don't tear down a fence before you understand what it does. As tempting it is, we can't rush to do away with blind review. Disclosing the reviewer names may also cause problems. The authors are partial about their work, and some may unfairly retaliate to the reviewer. Secondly, without the anonymity cloak, the reviewers may not be critical enough of some submissions from certain high-profile faculty/schools.

So with due caution and awareness of the complexity of peer review, I provide some suggestions to improve the system in good faith. This proposal has holes and difficulties in implementation, but I hope it serves to start the conversation.

While I am not confident in the specifics of the implementation, I think the high level solution is clear. We need to incentivize the reviewers and then demand accountability from them. So here are my suggestions to this end.

1. Compensate the reviewers

Reviewing is a thankless job. Conference reviewers get some recognition, as their name appears in the program committee page of the conference (but this gets old quickly). Journal reviewers don't even get that.

The reviewers do these reviews pro bono and are overworked. You may not realize it from outside but academics work really hard between teaching, research, departmental administration duties, and other professional duties.

Did I mention that reviewers and Editors don't get paid a dime? We are talking about domain experts here, who could easily make \$200 an hour consulting. I think we owe it to the reviewers to compensate them for their contribution in the publishing process.

Can the conferences and professional organizations afford to pay reviewers? I don't see why not. I have not been involved in any conference organization, so I may be wrong here, but please humor me.  A conference with 1000 attendees with \$1000 per registration fee (not uncommon) makes a total of  \$1 million. Where does this money go? Definitely not to the dubious quality conference swag. Hotels may be expensive, but not that much. If hotels are taking most of that money, we need tougher negotiators on the conference organization committees. The hotels already get a good deal with attendees staying there for the duration of the conference.

Remember Chesterton's fence. We should of course consider what type of side-effects compensating the reviewers may have. Could this lead to a clique of friends who recruit each other for reviewing gigs? I don't know.  If the compensation is not high, and if we keep reviewers accountable with respect to the quality of the reviews, this may not be a problem. If money is too messy, provide free conference registration to the reviewers.

Even if we don't compensate the reviewers, at least we need to set up a system to prevent freeloading. If you don't help with reviewing, you don't get to have your papers reviewed. If you are a conscientious reviewer, maybe you get to have a fourth reviewer?

2. Provide an author-feedback phase

Conferences should provide a rebuttal phase to give the authors a chance to respond to the reviewers' criticisms.  Although limited in their effectiveness/consequence, this response phase still gives a voice to the authors. As an extra point, I really liked what SIGMOD did with their author feedback; they explicitly asked the authors to report on any bad/offending reviews.

In journals, there is no rush. So even for rejected papers, the journal may provide a response opportunity, and the authors get to present their responses to the reviewers.

3. Form a grievance committee

To give more voice/faculty to the authors, a grievance committee can be formed to inspect the reviews suspected of foul play. The committee shall inspect the situation, consult with the other reviewers on the paper, and write a report on the decision.

Maybe this is on the crazy side, but here it is: It may even be possible to publicize the name of a malicious or severely negligent reviewer. (There may be a decentralized cryptographic signing solution under which two reviewers may make the name of the third reviewer visible if they agree on neglect/abuse by the third reviewer. Crypto geeks, is it possible to implement a multisig solution for this on hotcrp soon?)

4. Take ownership/responsibility for the reviews 

As PC chairs or journal editors, you should take responsibility of the quality of the reviews provided to the authors. You should not blindly sign on the reviews, as at the end of the day the quality of the reviews provided to the authors is your responsibility.

In addition, in-person PC meetings (when feasible) is good for enforcing accountability for the reviewers. Again the travel for PC members should be paid for by the conference registration fees, if an in-person PC meeting is established.

Reviews can be rated by other reviewers and feedback in the form of blind review can be provided to the reviewers. These feedback can help train reviewers to avoid common mistakes: being hypercritical on a minor point, failing to see the overall value provided, being prejudiced towards certain methods/approaches, making unsupported claims, etc. We may even consider pair-reviewing to complement peer-reviewing.

Finally, as a reviewer, you should consider voluntarily signing your names on your review. The idea here is to keep yourself accountable by voluntarily giving up your anonymity. The signing of the name decision should be made before the review assignments and one should not be allowed to sign reviews only for acceptances.

I have seen some people do this. And I will give this a try myself. In my experience, a reject decision doesn't hurt if the reviewer supports her position well and put in the work to understand and fairly evaluate the work with respect to the cohort of papers submitted. So I am OK signing my name on a reject decision.


Popular posts from this blog

Graviton2 and Graviton3

Foundational distributed systems papers

Learning a technical subject

Learning about distributed systems: where to start?

Strict-serializability, but at what cost, for what purpose?

CockroachDB: The Resilient Geo-Distributed SQL Database

Amazon Aurora: Design Considerations + On Avoiding Distributed Consensus for I/Os, Commits, and Membership Changes

Warp: Lightweight Multi-Key Transactions for Key-Value Stores

Anna: A Key-Value Store For Any Scale

Your attitude determines your success