Scientific Peer Review Is Broken – Fighting to Fix It with Anonymity
wired.com
wired.com
Ironically, the very title of this article is a model of non-scientific thinking that ought to be subject to criticism, but attempts to inoculate itself against criticism by asserting that the subjects of the piece are protagonists "fighting to fix" a broken system. They're not. They're fighting to replace one broken system with a different broken system.
There is no question that scientific peer review is broken and needs to be fixed. But anonymity is not the answer. And holding anonymity up as something that should itself be beyond criticism is certainly not the answer.
Of course, the opposite can happen where mediocre works slips through, and the reviewers that allowed that should be held accountable too. Its painful to me that so much of the acceptance process for research papers (in my field at least) is based on luck.
Moral of the story: reviewer anonymity is good but it comes at the expensive of accountability.
> Of course, the opposite can happen where mediocre works slips through
Slips through? Mediocre work is the norm, or at least close to being the norm. (And, as you say, good work is often rejected.)
> Its painful to me that so much of the acceptance process for research papers (in my field at least) is based on luck.
I don't think it's based on luck, it's based on politics. Academic science is a racket. There is some luck too, though.
In computer science, this is likely because of the ridiculous number of conferences there are and how eager each one is to fill out their roster. I want to hear of the conference that published no papers this year because none were good enough for publication.
> I don't think it's based on luck, it's based on politics. Academic science is a racket.
I've heard (more than once) of people who submit a paper, have it rejected, and then get scooped by a professor who was on the PC of the conference that rejected them. There is a very real possibility of work being accepted or rejected purely because of what the reviewer thinks will be most beneficial to their career.
So, I think your cynicism is well-placed.
Yep. But there is a reason it is this way, which is that to be successul on the academic feeding trough/career path, you have to serve on PC committees and you have to publish a lot of papers.
> There is a very real possibility of work being accepted or rejected purely because of what the reviewer thinks will be most beneficial to their career.
I think this is right, but it's less often so work can be scooped, and more often about building up your "camp." As a research professor, you need your area of expertise to be popular and you need to get a lot of citations for your work. You want to be a leader in an area that actually matters, where your definition of "actually matters" is actually purely herd mentality: what do other people consider to be important? So you want to reject work that bolsters your camp, cites you, etc., and reject work that poses a threat to the ascendancy of your area of expertise.
Minor reform of the referee system is treating the symptoms, not the disease.
The disease is that there is no honest, mutual, voluntary exchange between a party that values a given research project and the party that produces it.
Rather, we have federal bureaucrats handing out money taken from taxpayers willy-nilly via an old-boy network as described above. (The difference between taxpayers and slaves is that taxpayers only must sacrifice a fraction of their productivity to this absurd system, not all of it.)
In CS, all the research we do either goes to benefit shareholders of companies that ultimately profit from it, or (the majority) is just ignored because it's part of the paper mill competition. Those companies should be the ones paying for the research, not the taxpayers. Those companies are free riding, and "we the people" should put an end to it.
> Unless you're proposing to eliminate the concept of publishing research at all...
I think putting my research on my website would be just as good as publishing it in a conference proceeding. I'm not against sharing research and having it be public, but I think conference proceeidngs and journals in CS are of very little value. If people want my research, they can find it on my website. And I think this generalizes... if I want your research, I can (presumably) find it on your website.
Ultimately, it may be useful to have some kind of aggregation of what research is coming out in various subfields. Internet fora and the like would serve the purpose fine. But sure, having a group of referees and some way to screen research and highlight what is good could certainly be useful. All that stuff should develop organically as it is needed, though.
The model we have now, with peer review, journals made out of dead trees, etc. is a hold-over from a pre-Internet time when a few people in every sphere of life controlled the information that was disseminated, because it wasn't possible for things to be open, since we lacked the technology. That is no longer an issue. Of course, academics will hold onto their little racket as long as they can (probably indefinitely).
In fact, even after so much work to get it published, we now know of two errors in it - neither were pointed out by the reviewers of course.
I can't help thinking that that one experience sums up all of your problem cases above :)
I think the appeal for the case to drop is not to discourage discussion of anonymity in review, but that it isn't discussion. A jury panel that decides to outcome of a paper could make science worse off than through a panel of peers.
Someone very clearly was trying to get Dr. Sarkar fired. The PubPeer comments are only one small aspect. But the person then also (I assume anonymously) emailed his new employer to allege fraud, and then went so far as to print out these allegations and stamped them with official looking nonsense about being from the an NIH investigation (which didn't exist), and distributed them throughout Sarkar's department mail boxes.
The complaint makes a decent case for why they think pretty much all the negativity directed at this researcher is likely from one angry person. They obviously can't prove that multiple anonymous comments are from the same person, but reading the content it certainly seems likely. What is assumed to be the same person then took things way past the line of what most of us would consider ethical.
There may very well be a place for anonymous calling out of potential research misconduct. But making fraud allegations anonymously online, then printing out those comments, trying to fake them to look like an official government inquiry, and physically delivering them to the researcher's boss at his place of work isn't the way to do that. This case might be one of those cases where the anonymity should be protected at all costs out of principle, but it's a really shitty case to wave your "we're the good guys" flag for.
The plausible evidence of persistent research misconduct in this case seems to be a reasonable basis for someone not to get a new job with more research responsibilities, so there may not be any actual harm to the plaintiff here. The defense of truth is strong in a defamation case.
As far as I know, I don't know any of the people involved in this, but I've read enough legal complaints to know that they're not always as factual as they first appear, so I try to consciously restrain myself from simply believing it without hearing the rebuttal. Rushing to judgement may be quite common these days, but it's not the sort of behavior fair or intelligent people should engage in.
It's weird that they lead in with the first few comments, which were tame enough that I discount them completely as being an appropriate subject for a lawsuit, though they seem to get progressively worse and some of the later allegations are much more serious. I do wonder why they lead with their weakest cards, though.
That aside, it does seem that the site it exercising editorial control over the comments, which may work against it with respect to the CDA, even though it's trying to do the right thing based on its policy as quoted in the complaint. Given the policy, I wonder how the later comments ever went through, as it seemed like the site wanted to avoid drama like this entirely. Even if this was a lapse of some kind, it might work against them.
As I have not heard the rebuttal, I will refrain from making any judgements one way or another about who is right here. It's only fair, after all.
EDIT: Retraction Watch is covering this case. I believe them to be a reliable, respected and neutral 3rd party: http://retractionwatch.com/2014/12/10/pubpeer-files-motion-d...
This is pretty factual discussion, and the fraud allegations are very convincing. Look at the pictures of the gels. (I don't understand why people use Photoshop to create fake images. You could create a fake gel in 30 minutes, make a real photo of it, and noone will ever be able to prove it.)
If I made a discovery like this, I'd also get pissed. I'd probably also contact the employer of this person.
http://retractionwatch.com/wp-content/uploads/2014/12/2014.1...
As an aside, I'll mention that the PubPeer lawyer is doing a much better job in filing pleadings, perhaps because PubPeer is a more above-board client, than Sarkar's lawyer is doing. Yeesh, what a mess the Sarkar complaint is. (Yes, I am a laywer, and I have read pleadings before and have a gut sense of which pleadings are convincing.) Note that the American Civil Liberties Union signed off as lawyers for PubPeer.
Due to the ferocious competition for grant money, people are either sloppy and cut corners, or do whatever they need to do and outright cheat to publish in top tier journals.
As a society, we need to make a decision about how much we want to fund scientific research - then, once we've made that decision, insure that we put in place a sustainable system in place - we cannot put in a put of money to finance 100 grants, but then build a pipeline that funnels an ever increasing amount of people into a pool that remains constant.
All things being equal, such approaches are generally discouraged by economists. So presumably you think that the market for research labor is unusual, e.g. has some market failure. Could you describe what you think that is?
Basically, my angle is that, once you've answered this question, it's almost always better to try to fix the market failure directly rather than go full command and control. For instance, if you think that prospective PhD students aren't well informed about the long odds against getting a tenured position, then it would be better to educate those students rather than instituting a cap on the number of PhDs that are funded.
One argument is that review of results is easier for a reviewee to defend in public, since the data and results are stated and there's nothing to hide. Reviewing funding proposals is more speculative once you get past basic methods or ethics review (e.g. depends on the risk appetite of the funders) and therefore might require anonymity.
This is something we've been grappling at Experiment. It's possible a better design of the system is a double blind review for proposals but a public post-pub review for results.
Identifying reviewers is a little harder, but it's usually not difficult to make educated guesses unless they are, as you put it intentionally vague. Otherwise the other comments on this article about bad reviews from 'people from "a different camp" simply disliking a given approach' and the like would be impossible to justify.
This model relies on t a blind cryptographic signature.(https://en.wikipedia.org/wiki/Blind_signature)
PeerPub generates a public/private key pair.
1. Alice wishes to register on PeerPub. She generates a Nonce N.
2. She blinds the nonce with a factor B.
3. She submits the blinded nonce and her identity to PeerPub. PeerPub checks her credentials and executes a blind signature of the number and returns it to Alice.
4.Alice now separately registers an account perhaps using a privacy protecting system like Tor. She uses the original N and the unblinded Signature.
5. PeerPub verifies N + signature and registers the account. PeerPub will have no way of linking N to the original credentials. PeerPub can record N and make sure it can only be used once to register an account.
http://private-idp.appspot.com/
https://code.google.com/p/pseudoid/
http://blind-signer.appspot.com/
I haven't touched it for years and it looks like the login flow is broken now with the demo sites.
Also, this uses unsafe server-side JS crypto for demo purposes only.
There are people who are thinking of multiple solutions, one that I think is interesting is the proposition of Open Scholar, dubbed "Independant Peer-Review". I submitted it to HN so we can discuss it without flooding this thread, if it interests HNers: https://news.ycombinator.com/item?id=8734271.
In there scheme, the papers are open access from the beginning, the peer reviewing process is open, and reviews are citable. I believe it would encourage better and deeper reviews.
It would also be very nice for young scientists or students who want to apply for a PhD grant to be able to show that they are able to write a comprehensive review of a paper in their field, and that their review was good enough to significantly improve the paper or to be selected by the authors and/or publishers to be released alongside the paper, for instance.
The problem with this mechanism is that reviewers have no liability, because their comment is anonymous to the author and won't be available to the readers, as it won't be published as part of the article. The result is that reviewers are not made accountable now or in the future for inaccuracies in their review, blatant attacks, or tactical requests for additional irrelevant investigation just out of spite or to stall you so that they can scoop your paper.
Occasionally, the Editor can step in and disregard a particularly obnoxious reviewer, but it depends on the editor, the journal, and the political/scientific strength of the reviewer.
You can make comments on your papers available; there is nothing against that. I really think, however, peer review comments should actually be published with rejected and accepted papers (at the author's consent, not reviewers) so that conferences can be more transparent.
On the other hand, should anonymous commenters have the balance of power: in other words, say whatever they want with impunity, even if it actually is defamatory?
(Not saying that is the case in this situation, but in general).
The problem is that defamation is a legal concept, which can only be tried legally. So for instance, whereas a site can rigorously enforce rules which say that all comments are directed at the research material, and not at persons, and have a factual basis in that material, those measures cannot take away the right of someone, who feels they have been defamed, to take the matter to court (where they will almost certainly lose, which is neither here nor there).
You can't just create a site and declare it above the law, so to speak.
The only way to protect the identities of the anonymous is for the site to take responsibility for the statements it publishes: to assert that the anonymous statements are subject to rigorous standards of review, and when published, they in fact reflect the views of PubPeer, and PubPeer alone, and not of any anonymous persons (who do not publish any statements, but act only as sources of information).
Then if someone feels they have been the target of defamation, the defendant shall be PubPeer.
> (1) the plaintiff must make good faith efforts to notify the poster and give the poster a reasonable opportunity to respond; (2) the plaintiff must specifically identify the poster's allegedly actionable statements; (3) the complaint must set forth a prima facie cause of action; (4) the plaintiff must support each element of the claim with sufficient evidence; and (5) "the court must balance the defendant's First Amendment right of anonymous free speech against the strength of the prima facie case presented and the necessity for the disclosure of the anonymous defendant's identity."
In case of defamation, there's already a pretty high barrier for prima facie cause of action, and relatively broad protection for anonymous speech.
There's also little special about online communications. Anonymous speech, and the difficulties in balancing the different factors, have existed since before the founding of the country, when anonymous pamphlets like 'Common Sense' were used to press the call for independence.
Regarding "the site [must] take responsibility for the statements it publishes", this is wrong. Otherwise HN must be responsible for everything posted here, and we would have no HN. The relevant law is Section 230 of the Communications Decency Act of 1996, which grants pretty broad immunity to service providers. See http://en.wikipedia.org/wiki/Section_230_of_the_Communicatio... .
So: 1) it's well-established that anonymous commenters cannot say whatever with impunity, 2) defamation in this case is little different than the many previous cases on the topic, 3) no one seriously believes you can create a site and declare it above the law (assuming no jurisdiction problems), 4) existing case law is on the side of PubPeer, and 5) CDA gives PubPeer the ability to host anonymous comments without automatically being the target of a defamation suit.
For safe harbor here, you'd be looking to the CDA, I believe. In that case, what matters is the level of editorial control you exercise over the site. The complaint alleges that they review every comment before posting, which if true, would seem to be a problem for them. My understanding of the CDA is that, somewhat perversely, you're better off having low standards than high ones with respect to user-submitted content.
That aside, the above is a layman's understanding of the relevant law, not legal advice. If you're ever in a situation where this information is important, please discuss it with a lawyer.
When people are anonymous, they ARE more likely to be truthful in their criticism. They have less incentive to hold back, and it's just human nature to tone down critical feedback when you're critiquing the work of someone with is either influential or an acquaintance. No one likes to make enemies.
On the other hand, anonymity can pretty clearly bring out some of the worst in us. Some people feel little obligation to be fair or honest when their reputation isn't on the line, and so you see people trying to knock down rivals, people they don't like, or random strangers just for the "thrill of the troll."
Imagine if every time you applied for a job your potential employer had access to anonymous feedback on your past work. Some of it might be fair and honest (whether positive or negative), but some of it might be lies from an anonymous coworker with a grudge. Maybe someone is trying to take you down a rung because you got the promotion over them. You could be penalized for any petty reason, and it would stick with you.
Anonymous feedback communicated publicly is much the same. It holds the reviewer and the object of review on unequal footing. Anonymous feedback would be great for an author or even an editor, but it's just not fair to allow the pettiest of people to attack the works of others while wearing a mask. I'd like a system that helps researchers and others invested in the work to solicit anonymous feedback to make the work better. Public-facing commentary, on the other hand, should be tied to an identity.
Farming out low-level tasks to automated systems would be interesting. Imagine if the format of papers changed entirely, ie, you had to submit your proofs in certain formats, or at least certain parts in specific ways. I'm sure that many professors would be elated to see the number of papers they have to review go down drastically; although, I'm sure many will be disappointed to get a return letter that says, "I'm sorry, but the low level flaws were so serious that they were automatically rejected and are not fit for review."
> Have you ever questioned the claims that scientists make? For example, last year’s discovery of the so-called “God particle,”
Using the nickname for the Higgs in the context of questioning science claim is nothing but bait for the foolish and misinformed, and in this way anyone scientifically literate should seriously question any claims or opinions in this publication.
I don't think this is something one should issue a lawsuit over--but I also don't think that their proposal of completely anonymous review is at all useful.
Our approach is to focus on turning review of all kinds (including both pre- and post-publication) into a measurable research output -- something you can add to your resume. We support both anonymous and signed review with the idea that it will lead to greater transparency in the long run and also motivate reviewers to contribute more.
We have a significant number of both types of review now and are starting to look ways to measure if there are significant differences between blind and open review.
I would think that if you expect people to take the "credit for reviewing" model seriously, then likewise they should take seriously the notion of "credit for being reviewed" or "credit for being upvoted".
Also, there would have to be a way to verify submitters/reviewers credentials (e.g. connection to a research institution). Obvious arXiv has a model for how this could be done.
Disclaimer: I'm a physicist. I'm sure that physics is not beyond critique, but unless somebody is willing to take enough of an interest in the inner workings of physics research to say something specific about it or credibly include it in a generalization, I'd rather say something like "medical peer review is broken."
I don't understand how "The government cannot outlaw your speech" means "Your anonymity is protected by law from private parties". How does that interpretation come about?
> "the court must balance the defendant's First Amendment right of anonymous free speech against the strength of the prima facie case presented and the necessity for the disclosure of the anonymous defendant's identity."
What is really being discussed there is if the government could compel Yahoo to unmask an anonymous user. It says nothing about actual anonymity guarantees among private parties. If Yahoo decided to hand over the account details, the court would not have intervened at all. So when you have an account with some service, any expectations on how the company will handle the data come from the TOS.
In a similar fashion, a company is free to delete any content you submit to their website. The government can't curtail your speech, barring a few exceptions, but a company certainly can, as long as the government is not involved.
"A prominent cancer scientist, unhappy with the attention his research papers have received on PubPeer, is suing some of our anonymous commenters for defamation."
That's a third-party, who is not a government organization, suing PubPeer to de-anonyomize a commenter on PubPeer's site.
"Dendrite International, Inc., a purveyor of computer software used in the pharmaceutical industry, brought a John Doe lawsuit against individuals who had anonymously posted criticisms of the company on a Yahoo message board. When Presiding Chancery Judge Kenneth MacKenzie rejected one of Dendrite's requests to compel Yahoo to reveal the identity of an anonymous defendant, Dendrite appealed."
That's a third party, Dendrite, who is not a government organization, suing Yahoo to de-anonyomize a commenter on Yahoo's site.
If Yahoo wanted to, they could turn that information over. The question was, could Dendrite compel Yahoo to do so if Yahoo didn't want to. And the answer is that there's a First Amendment issue which must be considered, so the answer is "sometimes."
Yahoo didn't want to. And PubPeer doesn't want to.
So I don't understand why you are bringing up first party actions (actions that Yahoo and PubPeer are free to do) when the issue is what third party actions may compel them to do.
(A personal example - I have just spent a week constructing a counterargument to a reviewer who didn't read my paper properly. Imagine having an online discussion where your career success hangs on the response of an anonymous, disengaged flamer to a single post of yours: can you imagine how much effort you put into that post? Sadly this particular argument is pointless and serves nobody; it would be better use of my time to get on with actual research, but that's not how the system works).
If there's anything we've learned by now about "crowdsourced" review systems, it's that, without an elaborate way to evaluate reviewers, they fail. Badly. Facebook "likes", Google "+1", and Yelp reviews are heavily spammed. This just does not work.
A better solution might be to enforce non-anonymous peer-reviews that can be read by the public after the paper has been published.
Some of the other comments here suggest that anonymity of reviewers is dangerous in itself. That's why some researchers promote an open review process. Jelte Wicherts and his co-authors put a set of general suggestions for more open data in science research in an article in Frontiers of Computational Neuroscience (an open-access journal).[2]
"With the emergence of online publishing, opportunities to maximize transparency of scientific research have grown considerably. However, these possibilities are still only marginally used. We argue for the implementation of (1) peer-reviewed peer review, (2) transparent editorial hierarchies, and (3) online data publication. First, peer-reviewed peer review entails a community-wide review system in which reviews are published online and rated by peers. This ensures accountability of reviewers, thereby increasing academic quality of reviews. Second, reviewers who write many highly regarded reviews may move to higher editorial positions. Third, online publication of data ensures the possibility of independent verification of inferential claims in published papers. This counters statistical errors and overly positive reporting of statistical results. We illustrate the benefits of these strategies by discussing an example in which the classical publication system has gone awry, namely controversial IQ research. We argue that this case would have likely been avoided using more transparent publication practices. We argue that the proposed system leads to better reviews, meritocratic editorial hierarchies, and a higher degree of replicability of statistical analyses."
Wicherts has published another article, "Publish (Your Data) or (Let the Data) Perish! Why Not Publish Your Data Too?"[3] on how important it is to make data available to other researchers. Wicherts does a lot of research on this issue to try to reduce the number of dubious publications in his main discipline, the psychology of human intelligence. When I see a new publication of primary research in that discipline, I don't take it seriously at all as a description of the facts of the world until I have read that independent researchers have examined the first author's data and found that they check out. Often the data are unavailable, or were misanalyzed in the first place.
[1] http://retractionwatch.com/2014/12/10/pubpeer-files-motion-d...
[2] Jelte M. Wicherts, Rogier A. Kievit, Marjan Bakker and Denny Borsboom. Letting the daylight in: reviewing the reviewers and other ways to maximize transparency in science. Front. Comput. Neurosci., 03 April 2012 doi: 10.3389/fncom.2012.00020
http://www.frontiersin.org/Computational_Neuroscience/10.338...
[3] Wicherts, J.M. & Bakker, M. (2012). Publish (your data) or (let the data) perish! Why not publish your data too? Intelligence,40, 73-76.
Let valued scientific reviewers gain reputation in a similar way to how contributors to HN, SE and other sites do.
There could even be a viable business model in this. Publishers, be creative!