Bad Science

September 26, 2016 11:52 PM

The Inevitable Evolution of Bad Science "Now, imagine you’re a researcher who wants to game this system. Here’s what you do. Run many small and statistically weak studies. Tweak your methods on the fly to ensure positive results. If you get negative results, sweep them under the rug. Never try to check old results; only pursue new and exciting ones. These are not just flights of fancy. We know that such practices abound."
posted by dhruva (15 comments total)

Yeah, the incentives in this profession are not good for academic research. We will always be judged on our publication history and publish or perish is the unofficial motto of research in the modern era. I agree with most of the article, although the commentary about statistical power - while on the mark - neglects to mention the ethical hurdles to using large numbers of animals for experiments. Not insurmountable, but definitely a contributing factor.

Beyond Retraction Watch and a few lone individuals, there is not a lot of enforcement out there to help ensure good science. As I see it, there are three elements that make up the thin blue red white line between good and questionable data: Individual scientists’ fear of being wrong; the desire to see clinical and commercial outcomes (which set much higher standards); and the fact that research and technical assistants, who are employed by scientists but slightly removed from the massive incentive to publish in prestigious journals, do tend to be assessed on the quality and robustness of their work, rather than whether it is novel or exciting.
posted by kisch mokusch at 12:15 AM on September 27, 2016 [3 favorites]

For anyone interested in this, I highly recommend the books "Bad Science" and "Bad Pharma" by Ben Goldacre (the latter in particular deals with many of the same issues in this article). As a lay person with no scientific education beyond GCSEs I found it absolutely fascinating as well as being very clearly explained.
posted by the long dark teatime of the soul at 5:08 AM on September 27, 2016

Hmmm....
posted by Fizz at 6:23 AM on September 27, 2016

When I first encountered this critique almost a decade ago it was called the winner's curse, like in auctions where the winner is always the person who overpays.

I find it disheartening that we keep on pointing out the problem and everyone nods sagely but returns to their bad science.
posted by anotherpanacea at 7:57 AM on September 27, 2016 [2 favorites]

I find it disheartening that we keep on pointing out the problem and everyone nods sagely but returns to their bad science.

"It is difficult to get a man to understand something, when his salary depends upon his not understanding it!" - Upton Sinclair
posted by splitpeasoup at 8:12 AM on September 27, 2016 [6 favorites]

The first principle is that you must not fool yourself — and you are the easiest person to fool. R. Feynman
posted by bukvich at 8:18 AM on September 27, 2016 [1 favorite]

A Rather Large Pharma set up a reproducibility group whose job was to replicate publications before chasing after a new cancer target. Less than 50% of the publications were usefully replicated.
posted by benzenedream at 8:23 AM on September 27, 2016 [1 favorite]

A Rather Large Pharma set up a reproducibility group whose job was to replicate publications before chasing after a new cancer target. Less than 50% of the publications were usefully replicated.

There are lots of issues conflated in questions of reproducibility including perverse and good incentives, the striaght-up- difficulty of some science, known-unknowns vs unknown unknowns, hype, misinformed readers and fraud.

Rather than rehash them here, please, as always, read thoughtful discussion here and here.

tl;dr, you must r!
posted by lalochezia at 8:42 AM on September 27, 2016 [1 favorite]

Part of the problem is the quasi-religious belief that Science is in the business of generating Truth, and it does so through Publication. Science is an activity that must constantly renew itself. Far from being a Truth-generator, it is a facilitator of consensus by discussion, argumentation and evidence that must be witnessed.
posted by stonepharisee at 11:36 AM on September 27, 2016

There are lots of issues conflated in questions of reproducibility including [...] the straight-up- difficulty of some science, known-unknowns vs unknown unknowns, hype, misinformed readers and fraud.

Agreed, and this is especially true in the social sciences.

But, I think that to say this is "difficulty" is misleading, or at least overlooks part of the problem. A lot of research, at least in psychology, looks at questions that don't have answers, or that can't be answered in a definitive way. I'm speaking specifically about all these attempts to make statistical predictive models of certain outcomes, or the relations between numerous variables in a social context. Why would we expect that social situations are simple/consistent/standardized enough to yield generalizable knowledge of things like "the most effective forms of educational media for children"??
posted by patrickdbyers at 3:32 PM on September 27, 2016

I don't know any academics who think that, stonepharisee. I don't think it's a major part of the problem. I mean, that is something that I expect from undergraduates when they first start reading research papers, not something from people more advanced in their careers. Is it different in your field?
posted by Kutsuwamushi at 3:35 PM on September 27, 2016 [3 favorites]

"Far from being a Truth-generator, it is a facilitator of consensus by discussion, argumentation and evidence that must be witnessed."

Just like all the disciplines ... except that rigorous reliance on confirmable evidence can produce highly reliable, testable predictions. If I have to walk across a gorge on a slippery log, I'd rely more on scientific analysis of my odds and best tactics than on consultations with an all-knowing oracle.

Any individual result is only as reliable as that individual. That reliance must be earned. Scientific consensus is not delivered easily; it's arrived at not in days or years but decades.

We humans have many ways to lean on "quasi-religious belief", and have long been encouraged to do so by those who will gain. Luckily we can avoid the dangers of that path, if we aren't pushed that way by corruption or greed or madness. Climate change is our prime example.
posted by Twang at 5:08 PM on September 27, 2016

The 7 biggest problems facing science
posted by dhruva at 9:34 PM on September 27, 2016 [1 favorite]

@Kutsuwamushi: I'm in cognitive science, a field polluted with unstated and unacknowledged reliance on concepts whose origin is theology (soul/mind) and not physics. Yes, there are others in my field who feel similarly.
posted by stonepharisee at 12:54 PM on September 28, 2016

Here's the original study discussed in the Atlantic piece. It's an interesting critique, and different from those mentioned in many of the comments. The idea is basically an evolutionary one. Novel results are more likely to get published, and bad science (such as studies with small sample sizes) is more likely to yield novel results. Given the pressure to publish, the upshot is that labs doing bad science are more likely to survive than those doing good science, because they produce more novel (and thus publishable results). So the whole system is essentially selecting for people who happen to do bad science.
posted by klausness at 4:47 PM on September 28, 2016 [1 favorite]