Positive Bias Could Erode Public Trust In Science

ananyo writes "Evidence is mounting that research is riddled with positive bias. Left unchecked, the problem could erode public trust, argues Dan Sarewitz, a science policy expert, in a comment piece in Nature. The piece cites a number of findings, including a 2005 paper by John Ioannidis that was one of the first to bring the problem to light ('Why Most Published Research Findings Are False'). More recently, researchers at Amgen were able to confirm the results of only six of 53 'landmark studies' in preclinical cancer research (interesting comments on publishing methodology). While the problem has been most evident in biomedical research, Sarewitz argues that systematic error is now prevalent in 'any field that seeks to predict the behavior of complex systems — economics, ecology, environmental science, epidemiology and so on.' 'Nothing will corrode public trust more than a creeping awareness that scientists are unable to live up to the standards that they have set for themselves,' he adds. Do Slashdot readers perceive positive bias to be a problem? And if so, what practical steps can be taken to put things right?"

Comments

Feelings are more important than science

[Gothmolly]

Right? isn't that what American schools and TV have been teaching for the last 30 years? Nerds aren't cool - facts are open to interpretation - everyone is special - you can eat more than you grow... When you have a society rewarding irrationality, what do you expect? Rigorous science?

Re:Feelings are more important than science

[AwesomeMcgee]

I wonder how many of these "positive bias" results come from the fact that if you publish results that disagree with the bias of those who are paying for the study, they'll probably ensure it's never published and you'll find yourself no longer running studies on their dollars.

In the tech industry we all deal with non-technical managers who drive the technical direction and often times define the message to the clients. Does science suffer the same unskilled managerial types pushing scientists to interpret results in a particular way perhaps?

I have a hard time believing a professional scientist doesn't know how to apply the scientific method, but then again incompetence is rampant in every other industry I guess, why not the scientific one..

Re:Feelings are more important than science

[next_ghost]

Positive bias works a little differently than you think. First, a few definitions. Positive result looks like this: "We tried X on Y and it works." Negative result looks like this: "We tried X on Y and it doesn't work." Which one looks more interesting? That's right, the positive one. Now remember that outside mathematics, science is stuck with probabilities. There's always a very small chance of false positives (and also false negatives, but those are less of a problem).

So let's suppose that we have an experiment which should come out negative with 99% certainty. There's 1% chance of false positive. Now let's have 100 scientists independently perform the experiment. Chances are that 1 scientist will get a false positive. This scientist will then publish a paper while the other 99 will give up and research something else without publishing because of the impression that negative results are not interesting.

This is how positive bias works. Those 99 negative outcomes need to be reported to show that the one false positive is a false positive but they aren't reported. It's not some kind of intentional fraud but merely a consequence of overt obsession with original and sensational results dictated by those who pay for research. It doesn't matter what the research actually means, only that it can be phrased as "We tried X on Y and it works."

Re:Feelings are more important than science

[radtea]

During most of my career in pure physics I got negative results, which were always hell to publish. One of the things you'd hear a lot was, "Don't worry, a negative result is just as good as a positive result!"

At some point I started telling friends and colleagues who got positive results, "Don't worry, a positive result is just as good as a negative result!" Which is false, as proven by the fact that no one but me ever said it.

Negative results are hard to get published, but far more common than positive results. Furthermore, on the road to any positive result there are going to be lots of negatives: even today, working in an area where true positives are much more common, I try to put a section in every paper entitled something like "Things That Did Not Work So Well", because any experiment or computation or theory is likely to involve some dead ends that seemed like a good idea at the time, and if scientists don't report on them they will continue to seem like good ideas to people who haven't tried them, who will then waste effort on trying them, and fail to publish them when they don't work...

Re:Feelings are more important than science

[matthewv789]

Yup, that's the crux of the problem. While it may be true, as others say below, that publication bias against negative results occurs in all fields (such as physics) regardless of study funding, what we are seeing now is the influence of pharmaceutical industry funding in the clinical trials used for FDA approval of drugs (that is, a company funding the trial of its own drug).

Specifically, drug studies funded by pharmaceutical companies are four times more likely to show a positive benefit than ones funded by neutral sources. This is a problem because nearly two-thirds of clinical trials used for FDA approval are now industry-funded.

Obvious Complex System

[Anonymous Coward]

Positive Bias is another word for Group Think. I guess it could also mean deception

data point

[Anonymous Coward]

I received my PhD in physics, and the thesis was measuring a number, in which I measured zero within the error bar. Not particularly interesting, but valid science. My wife was in a PhD program in Biology, she also did valid science, novel measurement technique, came up with an uninteresting result, therefore was not able to publish, therefore was unable to graduate. It would have been extremely simple to fudge the result to a 2-3 sigma result 'hinting' at an interesting answer, which would have gotten published. I think certain sciences have gotten to a point where they have forgotten that if you do valid work in a novel way, then that is science and you should not be punished for the conclusion of the measurement. Most measurements you do of the natural world should probably end up being unsurprising, and thus uninteresting, but you don't graduate or get tenure with those kinds of results. I think this is the mechanism for the positive bias. That is why I do not take results from certain branches of science at face value.

Positive bias is the wrong term.

[Karmashock]

They're overstating precision. Which rather then a forgivable error is an elementary mistake no trained scientist should ever make.

A VERY basic concept they teach at the lowest level of science education is the distinction between accuracy and precision. This is science 101.

Accuracy is whether or not a given conclusion is correct.
Precision is to the degree of specificity.

Typically you run into problems on complex subjects because they overstate the precision of their data or their ability analyze the data.

This can boil down to simple thinks like significant digits.

For example, I'm measuring volume to two significant digits in a giant data set with thousands of measurements. When and if I average those numbers the final average can't have more then two significant digits. That sounds elementary but you see this error made on some big studies. You'll have a situation where something is being measured in a crude sense by many sources and then in the analysis a much higher degree of specificity is implied.

Often that degree of specificity is required to make certain conclusions which is why they break the rule. This is lazy and a breach of scientific ethics. What they need to do is collect the data all over again this time to the level of specificity they need.

Simply saying its too hard to collect the data properly so they're going to make assumptions is not reasonable or ethical. I suppose you could do it so long as you kept an asterisk next to the data and the findings to make it very clear throughout that the conclusion is a guess and not in any way empirical science since at some point people were guesstimating results.

Re:Positive bias is the wrong term.

[ceoyoyo]

"When and if I average those numbers the final average can't have more then two significant digits."

Yes, the average can have more precision than the individual measurements. That's actually kind of the point of an average. It can't improve accuracy though, for the most common definitions of accuracy.

Your definitions of accuracy and precision are sort of right, but also sort of misleading. And the way you use specificity is incorrect. But as you correctly point out, a lot of working scientists are a bit fuzzy on all these concepts too.

Re:Wow! I guess Science HAS become a religion

[AwesomeMcgee]

Science is a method dingbat, anyone who puts faith in a scientist however is practicing demagoguery.

Practice science, not demagoguery.

Re:Wait, what?

[thegreatemu]

While I agree that models are frequently refined, leading to new results, there is a disturbing trend I see, not having to do with positive bias necessarily, but with uncertainty estimation.

One thing that I've found incredibly hard to beat into undergrads taking my physics lab courses is that getting your uncertainties (or error bars) right is far more important than getting the right central value. This is because uncertainties are the only way that two experiments can be compared against each other, or the only way to compare experiment to theory. If I have two models of climate change, one of which predicts a temperature rise of 3 C ± 5% and another that predicts 4 C ± 7%, those results are in large disagreement, whereas two studies that predict 20 C ± 15% and 40 C plusmn 35% are in much closer agreement.

But I see it seems much more frequently, especially in fields like astronomy, too little thought goes into the systematic uncertainties, and you'll get 4 experiments measuring the same thing with results that cannot be reconciled if you take their statistics at face value. This was a huge problem with many of the early global warming predictions as well; every year a new estimation would come out that was completely incompatible with the previous one. Yes, these models are insanely complicated, and it's damn hard to understand all the systematics. And of course you can't put in error bars for plain old mistakes. But do it too many times, and people begin to lose any faith that your estimates can be relied on for anything.

This is the problem I see; not necessarily bias toward a positive result, but a bias toward underestimating the uncertainty of your measurement, which I suppose could be different sides of the same coin. (E.g., a result of 2 ± 0.1 is a positive result; a result of 2 ± 5 is not!).

Science comes when results are confirmed

[Geoffrey.landis]

Actually, science is stll working; the real trouble comes with the publicity of the science.

You should never believe the results of any single study. Every scientist knows this; or should know this. Science comes when results are confirmed, not when somebody publishes the first paper. The real work of science just starts when somebody publishes a study saying "we show that x has the effect y." That initial paper really is no more than "here's a place to start looking." However, newspapers want to publish news, and they need to publish whatever's hot and interesting and being done today, not "well, scientist z had his team take a look at the xy phenomenon to see if there was anything interesting there, and they couldn't really find anything there, although maybe some other research lab might have different results."

And, I suppose that somebody should post a link to the obligatory xkcd: http://xkcd.com/882/

Re:Science comes when results are confirmed

[silentcoder]

There's a brilliant line on this in "The science of discworld" - I won't pretend I can quote it 100% accurately off the top of my head but it goes something like this:
"In the media you will often read that a certain scientist is trying to prove a theory. Maybe it's because journalists are trained in journalism and don't know how science works or maybe it's because journalists are trained in journalism and don't care how science works - but a good scientist never tries to prove her theory, a good scientist tries her best to disprove her theory before somebody else does it for her, failing to disprove it is what makes a theory trustworthy."

Re:Science comes when results are confirmed

[rgbatduke]

Actually, you need to read "The Black Swan" by N. N. Taleb. Science that tries to confirm a theory is already infected with confirmation bias. There are a pile of examples that demonstrate the fallacy of confirmatory inference. Taleb uses a variant of Bertrand Russell's -- a turkey might reasonably infer, based on his daily experience, that humans exist for the sole purpose of feeding him, caring for him, providing for his every need. This might go on for day after day, increasing the turkey's degree of belief in his hypothesis of a good and beneficent humanity filled with love of turkeys, right up to the day that -- ulp -- something unexpected happens.

I'm not a Popperite, rather a Jaynes-Cox-Bayesian, but nevertheless it is important to avoid confounding the relative strength of positive and negative evidence. Absence of evidence is not the same as evidence of absence, yet we almost invariably confound the two.

Taleb damn skippy agrees with you about publicity, however, and the near-criminality of publicity and reporting of science. A newspaper necessarily takes a scientific result or observation and transforms it two ways: First of all, it creates a narrative. It isn't just "a tornado hit Houston", but "a tornado hit Houston, possibly caused by Anthropogenic Global Warming" with the subtext "this isn't an act of nature, random an unpredictable, but is instead our fault". Aztec priests couldn't have come up with a better excuse for ripping the still beating hearts out of a stream of slaves and war captives. Second, it necessarily reduces the complexity of the result to no more than three variables, ideally one. It "Platonifies" it (according to Taleb) -- wraps it up in a pretty, easy to understand package that makes it more predictable, less random than it really was. Global warming is a much simpler "cause" than "A cold front overrunning a warm wet surface layer of air near the ground, creating turbulent rolls that break off and terminate on the ground, sustained and driven by the thermal difference, and it is a better story too.

Sadly, as you point out, real science is all too often (and should be) scientist z looked at something and didn't find much. But what they failed to find and how they looked is actually often as or more important than a study that claims to find something, especially when the latter uses questionable methodology to try to prove something, cherrypicks data (for the same purpose), ignores silent evidence (ditto) etc. Medical science is permeated with this. Nobody gets famous, or rich, or even a job, for looking for a cure for cancer and not finding one. This too is addressed by Taleb. Great book.

rgb

Re:Wait, what?

[jo_ham]

And there's jerks like you who claim that anyone who dare look askance at your work are "anti-intellectual" are are too stupid to sort out what to trust in a scientific publication (are you saying that scientific publications, Nature, et. al. have untrustworthy material in it?)

Ah, classic twisting of my words. I'm talking about anti-intellectualism as a movement. Those doing it are extremely smart - that's what makes them good at it. They're good at duping people into following their cause. I am not personally calling individual people stupid unless they actually demonstrate that they are.

And yes, I'm absolutely saying that scientific publications contain untrustworthy information, including the big hitters like Science and Nature - their size and prestige is no assurance of infallibility, and in fact can work against them since people are often reluctant to question them. That's the nature of scientific publishing - until results have been replicated, single-source experiments and models need to be looked at with extreme skepticism.

Last year I performed some work that disproved a piece of published literature (in a non-controversial area of chemistry). I didn't set out to disprove it - I set out to see if I could replicate the results and I determined that the published paper was incorrect. My own conclusions, method and data set were published in response, with some discussion on why the previous paper was drawing incorrect conclusions (mainly an issue with experimental control). My situation is one that is repeated constantly - it's how science works. The stuff that can't be replicated is corrected, the stuff that is replicated becomes more solid.

It's not the layman's fault that they don;t understand some of the intricacies of how peer review and scientific publishing and research works. They're not stupid for not getting that in the same way that I'm not stupid for not understanding the first thing about programming - it's simply not my area of expertise. Where the stupidity *does* arise, however, is when people start to distrust scientists out of hand because they're being told to do so by certain media outlets or special interests. It happened with vaccinations due to a corrupt doctor manipulating a very weak study with the ultimate aim to push a competing vaccine made by a company that paid him off, but it backfired spectacularly - far from getting the competing vaccine popular, people rejected vaccination entirely against their own interests, putting their own and everyone else's kids at more risk. It's this sort of media frenzy and associated public panic and distrust of science (even now, people refuse to believe scientists on the issue, despite the original study being totally debunked and Wakefield himself being struck off the medical register et and the whole thing exposed as a sham).

That's the sort of thing I'm talking about here. We saw it with the MMR vaccine, we see it with nuclear power, we see it with stem cell research, we see it with GM foods (and there *are* some legitimate issues to be raised there, being drowned out by typical media hysteria), we see it with climate science - again, there are legitimate issues to be raised and discussed on a topic that is *gigantic* in scope in the scientific community, but it's being drowned in so much media hysteria and political propaganda that it's almost impossible to get anything done.

Of course, equating people who don't drink your Kool Aid to those who deny the Holocaust really helps your cause to be seen as our nights in shining armor.

Where did I say that? You're dangerously close to Godwining the thread by trying to imply that I brought that up when I did no such thing. The hyperbole serves no one, it only makes your arguments look weak.

I'm not looking to be anyone's "night [sic] in shining armor", nor are most scientists. We just work on the science in our field and go where that leads us. If we wanted to be knights rescuing people I'd have joined the fire service or something. I became a scientist to ultimately help mankind and further our collective knowledge, but I'm no superhero or white knight.

A perfect example of what?

[Anonymous Coward]

To me it seems a perfect example of how NON SCIENTISTS have made the expectations that scientists are not able to live up to.

The report was basically "Hey, guys, this looks odd, but we've corrected for everything we could think of. You may want to see if you can replicate this".

Nobody else managed to replicate.

MEANWHILE the original posters were being told how they had it wrong. They changed their process to accord for this and retained some extra difference, despite this. Another attempt by another team showed no effect, so the original scientists CONTINUED to see what could be the cause if not FTL neutrinos.

They then found that the 50ns difference can be accounted for if the connectors were not tight.

NOT, at that time, that this was the cause of the discrepancy, but that

a) the couplings were loose NOW, but they didn't check THEN, could have loosened since then
b) an effect of this is a slight delay, enough to remove the difference they'd seen

All absolutely fine and no "positive bias" AT ALL.

But what is the public saying about it? What were the press saying about it?

THAT debacle (and how it's ignored by the parties involved in it) is a perfect example of the problems science has with people who aren't scientists.