Via Andrew Gelman’s site, here’s a TED talk by US philanthropist Laura Arnold entitled “The Four Most Dangerous Words? A New Study Shows”. It details her journey through the world of social, medical, psychological and other research. It’s a lively and concise summary of the developing critique of recent scientific research standards.
With her husband John, Laura Arnold funded The Center for Open Science, which sponsored the Reproducibility Project under the leadership of Brian Nosek. Nosek and the Center are central players in the emerging field of meta-research – research into better ways to do research.
In January Nosek, John P. A. Ioannidis and others published in Nature their “Manifesto for reproducible science“, an important roadmap for moving scientific research results closer to truth. It ends with one of the great Richard Feynman quotes:
“The first principle is that you must not fool yourself – and you are the easiest person to fool.”
David on Twitter: @shorewalker1
I laud their aims, but I am quite convinced their actions will have the exact opposite effect: all these ‘quality measures’ that research must satisfy will just shut out the competition and make it even more elitist and closed to outsiders than before, resulting in even stronger group think. They make the classic mistake that more regulations (overseen by the trusted few, ie themselves) mean better outcomes. You already see this in the pre-registration systems that are now emerging in social science, which enforces a long delay between design and outcome, effectively shutting out most undergrads and groups that lack deep pockets from the research game.
There is something inherently deeply unscientific about the notion of pre-registration: since many findings are in truth accidental, those findings are labelled ‘untrustworthy’ if you honestly report them. This forces you to be dishonest and not declare your new finding when you first make it, and then register another study setting out to find what you already know is out there. That will be the result of this kind of thought police: the real research will happen in secret, and then made visible via a pre-registered study, with no transparency about the actual process at all, which means others will make the same unseen mistakes, slowing down science.
Just ask yourself: from what crystal ball do the research hypotheses come? Where does the magic happen? Within the notion of pre-registration, the magic is a form of divine revelation wherein the brilliant researcher is struck by inspiration, after which he plans his experiments, registers what he expects to find, and then runs the study. Ha! If you believe science is really like that, you have been had.
So we are witnessing another round in a kind of purity-drive. Bureaucrats/Meta-scientists will love it. It will enslave real scientists and slow us all down.
I know an older saying than the Feynman one: the road to hell…..
Those guys themselves also appear to deliberately overstate the problem — in their study, they use what I would guess are two out of three of the worst high impact journals for p-hacking (Psych Science and JPSP), both of which are notorious for publishing whizz-bang effects designed to catch media attention but often without any great theoretical bases (JEP:LMC is better). This would be likely saying cars are all terrible because I found problems with many Ladas.
I also agree with Paul — I don’t see how pre-registration solves anything. I can just run a study, look at the results, and then pre-register them, and no-one would ever know. In some areas, people also copy each other’s studies sometimes (and hence publish the same thing first), and I can’t imagine too many people in these areas are going to go around telling everyone what they will do in exact detail in the next few years.
yes, the pre-registration system will quickly become a nightmare. Just imagine what people will pre-register when it comes to the analysis of a dataset with thousands of variables running for decades (like the cohort studies or the genetic surveys). They will start hoarding research areas by pre-registering masses of intended analyses, at great bureaucratic expense. And one can already see the plagiarism battles ahead where someone publishes something on an area pre-registered by someone else.
Yes, the holier-than-thou attitude by these self-appointed supra-scientists is simply going to add sand to the wheels.
Much better would be to let competition do its work: when bad and wrong ideas get published, then publish their rebuttal too and embarrass the original study as a deterrent. What is interesting about the supposed factoid that over half of the reported studies are not reproducible is that no-one bothers with most of them because rebuttals happen for the ones that matter. The rest is implicitly discarded as entertainment, not worth re-doing.
A topic after my own heart! Our RealKM online magazine has catalogued an range of discussions about trangressions in this space.
Paul and Conrad’s responses are instructive though. Your kneejerk responses (not meant pejoratively) demonstrate that the penalties of gaming the system are so inconsequential that people won’t think twice about bypassing any added controls.
To say that pre-registration is an added barrier is ludicrous unless you are already fully ensconced in the politics and incentives of the (broken) academic system. Why couldn’t pre-registration be as simple as emailing a friend and using that as proof of intent later on? Who cares if someone does a second study before the first is complete? Aren’t we trying to get more replication of work anyway?
It is highly unfortunate for the reputation and advancement of science that today, academics must focus first and foremost on being attributed to publicly acclaimed research. Without this they cannot maintain their personal standing, and thereby attain direct or indirect financial gain from their work.
Stephen, your idea about pre-registration is not how it works currently. The best such system at the moment is probably that of the Royal Society. At present, you send in your study, it then gets reviewed (which can take months), and then they are obliged to publish your article no matter what the results after you have got the data. So rather than just yet another mine to step on and to waste more time as you are suggesting (which no one will want to do due to pressure/bullying from their university management), they offer something useful in return — good ideas that give unexpected results are still worthwhile knowing about. I also like Paul’s idea, which in my field has more or less sprung up without any external encouragement– it is now common for labs to gang together to run the same experiment to get reliability, especially for contentious effects.
In terms of corruption and money, you are certainly wrong about the effect of money. The vast majority of academics don’t receive easy money and the majority will have no large grants in their career where they can spend the money. In fact, the vast majority of academics receive essentially no money. So the corruption comes from KPI management mania. If you try and fight this, you can ask Paul what happens!
Apart from this, instead of harassing the poor scientist at the end of the chain, who already has to put up with such a large and crazy bureaucracy making even the simplest thing hard to do, why not regulate the fund givers or the the university management who encourage taking from these companies who really science-advertising ? The latter of these loves nothing more than rules and forms. It seems to me your suggestion is analogous to the person who yells at the nurse when something in a hospital goes wrong. Sure, they may be involved in something that wasn’t great, but the real problems are usually caused from above.
Conrad,
Indeed, I fully agree that the problem stems from funders and institutions – “KPI management mania” as you say. Bad behaviour may be literally by scientists, but it’s certainly not initiated by them. It is a case of responding rationally to terrible/perverse incentives imposed by the institutions.
Sorry if you misunderstood, I’m absolutely not suggesting that the scientists themselves are receiving these rivers of gold.
Thanks for illuminating me about the Royal Society. I like the idea!
Stephen,
both Conrad and I have been involved in such debates for years, so don’t take the easy writing above as evidence of a knee-jerk reaction. I truly do think we are looking at a step in the wrong direction (and I do have solutions of my own to the problems with peer review, see eg. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2769202 ).
I think your label of “added controls” says it all, really. That is exactly the mindset of this mob. The idea is that another layer of self-appointed administrators will ‘help’ scientists stay on the streight and narrow. Just as the many layers of administrators is helping them already in Australia?
Nick has nicely laid out some of the issues on http://clubtroppo.lateraleconomics.com.au/2017/02/06/ethics-of-the-unethical-variety/ where he also links to other thoughts on such bureaucratic systems giving ‘added controls’.
Paul,
A significant problem with the bulk of current scientific research is the corruption caused by its corralling down paths which are more profitable / rent-seeking to the institutions that sponsor them.
I am not in favour of another layer of bureaucracy. But I am in favour of the scientific community articulating and encouraging what higher standards of experimentation look like, providing a peer counter-pressure to the financial incentives. Am I being too optimistic in hoping that there are still scientists out there who are genuinely seeking the truth?
I quickly read your paper. It’s a clever idea, but I can’t help but feel that this is a band-aid over a fundamentally busted scientific publishing process. First of all, it retains the peer review gatekeeper. Second of all, the problem is often not with the papers but with the communication that happens afterwards.
Even if the research itself is sound, the abstract will be written for maximum interest, and then the institution’s media team will devise a almost-but-not-quite lie based on that abstract.
The media, running to deadline, pick up on the strong implication in the media release, dropping the weasel words used by the institution that are there solely as protection against accusations of outright fraud.
Perhaps it is too much to expect that scientists promote science at the expense of their own jobs. If this is the case, what we really need is the equivalent of the ACCC, a watchdog which has some teeth in pursuing institutions who take public dollars without taking due care about the public benefit delivered — the real benefit, not the espoused hype-driven BS.
Paul have read that a problem for the peer review system is that, because there are so many papers being submitted , there are few people have the time to thoroughly read and examine every paper properly.
An example given was of a medical research paper on a new drug (resulting in clinical trials on real people). The findings of the research paper turned out to be based on fraudulent data and also on serious errors in statistics. However it took several people, hundreds of hours of work to tease out exactly ‘what was wrong ‘ with just that one research paper.
Stephen,
you say “I am not in favour of another layer of bureaucracy”
and a bit later
“what we really need is the equivalent of the ACCC, a watchdog which has some teeth in pursuing institutions who take public dollars without taking due care about the public benefit delivered — the real benefit, not the espoused hype-driven BS.”
it seems to me you are not quite honest with yourself: you ARE in favor of another layer of bureaucracy giving edicts to university management and grant institutions who will undoubtedly pass those on to the scientists. I can just see the forms that the ARC wants of Australian scientists getting even longer and more convoluted. They will just love your kind of argumentation.
I think it is futile to expect the free press not to exaggerate, no matter how careful the media release, nor do I think it is actually useful in social science to tell scientists to refrain from interpreting their results in such a way that others might get something out of it (which by necessity means translating into broader labels that others understand). After all, you want social scientists to be relevant, which by design requires extrapolating from their study which is always specific and backwards looking.
For the media too, you want the market to do its job and create quality newspapers that have a reputation for being more careful. You can have a national broadcaster that you want to be careful, but it is hard to see how you can prevent the free press from engaging in entertainment-by-exaggeration unless you go all-out and simply forbid a free press.
Paul,
I said “not in favour of”, not “would never support” — it is not inconsistent to argue that self regulation is preferable, while accepting the need for additional controls where systemic failures are occurring.
I doubt we’re going to agree, but I do see a difference between an institution like the ACCC which post-facto regulates company behaviour that falls outside accepted norms, and the ARC which creates a distorting environment to undertake (or not undertake) certain types of research.
From today’s HW in the Oz:
Its quite easy for peer review systems to ( gradualy) become clubs for the right kind of person, no?
yes indeed, the peer review system as it currently works is a conveyor belt for clique formation. If you put a regulator on top of that, such as a grant agency, the regulator becomes the new uber-clique. If you really want desinterested science, you have to stop funding it and rely solely on those enthusiasts willing to do it for nothing. You will quickly see most problems with peer-review disappear, but you would also see most research disappear….