Slashdot Mirror
How Science Goes Wrong
dryriver sends this article from the Economist:
"A simple idea underpins science: 'trust, but verify'. Results should always be subject to challenge from experiment. That simple but powerful idea has generated a vast body of knowledge. Since its birth in the 17th century, modern science has changed the world beyond recognition, and overwhelmingly for the better. But success can breed complacency. Modern scientists are doing too much trusting and not enough verifying — to the detriment of the whole of science, and of humanity. Too many of the findings that fill the academic ether are the result of shoddy experiments or poor analysis (see article). A rule of thumb among biotechnology venture-capitalists is that half of published research cannot be replicated. Even that may be optimistic. Last year researchers at one biotech firm, Amgen, found they could reproduce just six of 53 'landmark' studies in cancer research. Earlier, a group at Bayer, a drug company, managed to repeat just a quarter of 67 similarly important papers. A leading computer scientist frets that three-quarters of papers in his subfield are bunk. In 2000-10 roughly 80,000 patients took part in clinical trials based on research that was later retracted because of mistakes or improprieties. Even when flawed research does not put people's lives at risk — and much of it is too far from the market to do so — it squanders money and the efforts of some of the world's best minds. The opportunity costs of stymied progress are hard to quantify, but they are likely to be vast. And they could be rising."
FYI, 'trust, but verify' is also a great rule of thumb for spell-check.
Are the numbers from this article just pulled out of a hat?
All researchers in the sciences have a motivation to be published, in the form of recognition, academic progress, and financial motivation. Not many of them have an incentive stop working on looking great for producing results and check the work of someone else.
Assuming TFA's numbers are correct, I'd bet that much of the problem is that no agency, be it government or commercial (and particularly commercial) wants to spend it's money seeing if published results are reproducible. Additionally, no one ever won a Noble Prize for excellence in reproducing others' results. Verification of results is key to science, but this is one of several aspects of doing science right that the funding agencies either don't want to, or can't (as in Congress looking over the shoulders of managers at the NSF), pay for. Everyone wants "everything, all the time" without paying for it, and this is the sort of thing that happens when decisions are driven by the money people (who may be scientists, to be fair) and not the people who know what the hell is going on.
No. Actually it isn't. Because homeopathy can't even replicate their own results in controlled environments.
It SHOULD, however, be a wakeup call to scientists all over that their chosen fields are more caught up in the "publish or perish" mentality than they should be.
Between this, and others willing to take these unreplicated (and possibly unreproducible) studies as "Holy Writ", what people think of as science IS becoming as sloppy as religion.
Which makes it harder for the people who actually DO the grunt work and the follow-up to receive their just due.
Chas - The one, the only.
THANK GOD!!!
A big part of the problem is null results. Getting and reporting a null result is SUPPOSED to be good science. And in a lot of areas of science, its OK - you'd obviously prefer to find something cool, but you don't kill your career by not finding something. But in some fields, if you do a study that doesn't find something, you can literally set your career back a decade or. Guess what this leads to?
I was talking to somebody was getting her PhD in Biochem. She was in the midst of a 5 year study on the effects of some drug. A condition to get her PhD was that she must publish a "substantial" peer review result. And her department had gone out of their way to define null results as not substantial. This meant that if her study found that the drug wasn't effective, she didn't get her PhD - she would have literally had to start over and had wasted 5 years of her life.
This is common in some areas of science, but not others.
So the first thing to do is get rid of garbage policies like this. My understanding is that its much more common in biology related fields, but that might just be my bias (I'm from a physics background, so I have an admitted bias here).
Until you fix this policies like this, you will always have people getting "creative" with their statistics or just outright making up data. For some reason, a lot of these biology related fields don't seem to care about policies like this, which I just don't understand. I mean, we know that these policies lead to bad behavior, but nothing is done to fix it. Maybe somebody in these areas can explain the rational to me.
http://www.its.caltech.edu/~dg/crunch_art.html
"The crises that face science are not limited to jobs and research funds. Those are bad enough, but they are just the beginning. Under stress from those problems, other parts of the scientific enterprise have started showing signs of distress. One of the most essential is the matter of honesty and ethical behavior among scientists.
The public and the scientific community have both been shocked in recent years by an increasing number of cases of fraud committed by scientists. There is little doubt that the perpetrators in these cases felt themselves under intense pressure to compete for scarce resources, even by cheating if necessary. As the pressure increases, this kind of dishonesty is almost sure to become more common.
Other kinds of dishonesty will also become more common. For example, peer review, one of the crucial pillars of the whole edifice, is in critical danger. Peer review is used by scientific journals to decide what papers to publish, and by granting agencies such as the National Science Foundation to decide what research to support. Journals in most cases, and agencies in some cases operate by sending manuscripts or research proposals to referees who are recognized experts on the scientific issues in question, and whose identity will not be revealed to the authors of the papers or proposals. Obviously, good decisions on what research should be supported and what results should be published are crucial to the proper functioning of science.
Peer review is usually quite a good way to identify valid science. Of course, a referee will occasionally fail to appreciate a truly visionary or revolutionary idea, but by and large, peer review works pretty well so long as scientific validity is the only issue at stake. However, it is not at all suited to arbitrate an intense competition for research funds or for editorial space in prestigious journals. There are many reasons for this, not the least being the fact that the referees have an obvious conflict of interest, since they are themselves competitors for the same resources. This point seems to be another one of those relativistic anomalies, obvious to any outside observer, but invisible to those of us who are falling into the black hole. It would take impossibly high ethical standards for referees to avoid taking advantage of their privileged anonymity to advance their own interests, but as time goes on, more and more referees have their ethical standards eroded as a consequence of having themselves been victimized by unfair reviews when they were authors. Peer review is thus one among many examples of practices that were well suited to the time of exponential expansion, but will become increasingly dysfunctional in the difficult future we face."
I've collected some other quotes on social problems in science here:
http://www.pdfernhout.net/to-james-randi-on-skepticism-about-mainstream-science.html#Some_quotes_on_social_problems_in_science
A 21st century issue: the irony of technologies of abundance in the hands of those still thinking in terms of scarcity.
Really everyone should. Because while some of the points are good, The Economist misses the biggest one of all
In the University environment of today, the scientists and researchers are hamstrung by Non-Disclosure agreements. How does one share experimental information when to do so will cause you and your University great problems? One of the biggest offenders is the Biotech industry. Talk to someone, lose your funding and probably your job.
This is just the culmination of the past several decades shift from Government sponsored research to industry dominated research. It's a completely understandable position - industry wants return on it's investment, and research that doesn't generate profit might be good research, might be groundbreaking, but to the industry sponsoring the research it is a failure if they don't profit from it.
I'm pretty certain that industry would consider completely flawed and incorrect research as successful if it generated money for the company sponsoring the research.
So they draw the conclusion that scientists are lazy. I draw the conclusion that this is what happens when making money is the most important factor, and the scientists are bound by their contracts.
The shepherds did so well protecting the flock that the sheep no longer believed that wolves existed.
trust, but verify?
We must be talking about different sorts of science, because from what i know, the simple idea rather is
"be objective, and be sure to keep it falsifiable
CLI paste? paste.pr0.tips!
When I was at Cornell, I often talked to the people in the psychology department. One of the students told me she wanted to do an experiment that went something like this--it had been found by others that under certain circumstances, X, rats did something, A. She was curious as to whether, if she changed the circumstances to Y, they would still do A. So her proposal was to do the experiment under circumstances Y and see if they still did A.
I explained to her that it was necessary first to repeat in her laboratory the experiment of the other person--to do it under condition X to see if she could also get result A, and then change to Y and see if A changed. Then she would know the the real difference was the thing she thought she had under control.
She was very delighted with this new idea, and went to her professor. And his reply was, no, you cannot do that, because the experiment has already been done and you would be wasting time. This was in about 1947 or so, and it seems to have been the general policy then to not try to repeat psychological experiments, but only to change the conditions and see what happened.
Nowadays, there's a certain danger of the same thing happening, even in the famous field of physics.
"First they came for the slanderers and i said nothing."
If you are trying to prove to young-earth creationists that the earth is old because they should trust scientists, you're doing the wrong thing. If you do that, you turn it into a fight about, "the guy I trust" vs "the guy you trust."
Instead, if you really want to talk to a young earth creationist (I don't know why you would), you need to show them the evidence. Really dig deep. If they want to discuss carbon dating, then dig in and show the evidence we have of why carbon dating works. Eventually, if they are willing to go along with you (and it will take a lot of work so they might not), they will turn into an old-earth creationist.
And you will absolutely learn something along the way. Never turn the discussion into an argument about "the guys I trust" vs "the guys you trust" because that argument is never won, by either side.
"First they came for the slanderers and i said nothing."
If you would be so kind as to replicate the catastrophic prediction results -- oh, wait, those can't be tested?
We're in the process of testing them right now. We should have results in 50 or 100 years although chances are we don't have to run the full experiment to see where it's heading.
What makes you think the results on CO2 sensitivity are "out-of-you-ass guesses" rather than just an expression of the uncertainty of the results? Where have you seen a scientist that links every bit of bad weather to AGW? There are some non-scientists who may do that but that's not science.
It's not about being lazy. Feynman famously addressed this in his "Cargo Cult Science" rant in his Caltech commencement address given in 1974. (There's no recording AFAIK, that link is to someone reading the transcript).
He makes very good points: funding is for new results. Attempting to repeat another scientists published work is not a new result (unless you can't), and many places won't even allow you to try, unless it's something very sexy like observing the Higgs boson or something. It's an important structural problem, and it was worth calling attention to forty years ago.
There's no doubt that some unscrupulous researchers have noticed this and are gaming the system. The incentives to do so are particularly high in biochem.
Socialism: a lie told by totalitarians and believed by fools.
All my mod points for the ability to edit. Of course what I meant was:
The scientists are not the ones who brought politics into it.
I'm not sure what 'publish or perish' has to do with it.
I do research. I can get funding from NIH from a well designed, well reasoned approach to learn something new. What I can't get is funding to replicate some other researcher's finding.
I'd be happy to do replication work in addition to novel research, but it's a simple fact that no one will pay for salary of lab techs, lab equipment, or reagents in order to replicate something, even if I'm willing to donate my own time.
Possibly more importantly, pseudoscience is the articles worst nightmare.
The defensiveness now built into some fields (and here I'm thinking climate science), because of unrelenting, personal attacks does put important discussions like this into a defensive context.
And this is another bitter fruit produced by the anti-science industry, because these discussions are important to have. There are a lot of mistakes in science, but (seeming to me increasingly) there is also data falsification and fraud. [Retraction watch](http://retractionwatch.wordpress.com/) is a great website, but it makes sickening reading, and I suspect that it only scratches the surface.
I mean, sometimes, no fucks whatsoever are given. How that got past peer review blows the mind. And any of these.
Remember this letter to Nature (FFS!) pointing out that 70% of the papers in one of their issues didn't say what the error bar represented. How that got past the reviewers is mind boggling. Imagining how it got past the authors requires mental gymnastics. (Since the letter, Nature articles are much better, but Peer Review is not what is catching the errors).
So, lets talk about errors in scientific research, and lets talk about scientific fraud. It's important because its rampant, and despite that there are nutjobs seeing it in their peculiar light lets not be put off. This conversation needs to be had more often, because the problem is dug in at the highest levels of academic prestige.
Props to the Economist for bringing this up. I'd like to see this discussed in Cell, Nature and Science. And I'd like to see credible career protection for whistle-blowers.
Sorry, but you're just wrong. There are many problems in science where too much money is involved (medicine etc) but evolution is well documented, studied and proven over and over again. There is, as the other AC says, a "colossal wealth" of data, including DNA similarities, fossils and plain old observation. The only "myth" regarding evolution is that there is any controversy at all about it outside of a few religious zealots.
That and you don't often get PHD's, published papers, and prestige for trying to duplicating and test a published findings.
If something is so important that you feel the need to post it on the internet... It probably isn't that important.
It's resource intensive, but also just plain difficult. For example, publications are never a full description of an experiment, just the highlights. It takes a skilled researcher to fill in the gaps and then a second level of skill to accurately carry it out.
Looking at it from another perspective, ignoring scientific developments which are the result of inspired genius (which I would argue are rare), every new publication is the more novel and difficult work that has been conducted to date. If it weren't, it would have been done already.
So how can you expect someone else (who wasn't able or interested to carry out the work themselves) to immediately duplicate cutting-edge work based on an incomplete description?. It's a bit amazing that up to 50% of publications could be replicated at all.
"Trust but verify." I hate that phrase. It's used all over the place, but it is meaningless.
Interestingly enough it's a Russian proverb made famous in the United States by Ronald Reagan.
In any situation where verification is appropriate, trust is not.
In the specific context in which Reagan used it, it indicated that both countries would trust the other enough to begin dismantling warheads, but would verify each other's progress towards the agreed upon targets. It took trust to start the process, and verification would keep it going.
I guess people just don't know what "trust" means any more.
That's entirely possible. However, in this case I think it means you should generally trust that the experiment was done in good faith, however, if you need to actually use the results, you should verify them first.
Fanatically anti-fanatical
He makes very good points: funding is for new results. Attempting to repeat another scientists published work is not a new result (unless you can't), and many places won't even allow you to try...
"If I have seen further it is by standing on the shoulders of giants"
Even though funding is for new results, to get new results, these days you will almost always need to build on what has gone before. So while scientists generally don't attempt to replicate published results, if the work is important, someone will eventually think of a way to extend the work, and rely on it to build something else. At that point it will become obvious if the original research is flawed.
So good science does eventually win, it just can take a longer time than people would like to spot frauds. Science works, the last century is a testament to that.
You can never know everything, and part of what you do know will always be wrong. Perhaps even the most important part.
If you're a tech type (and I assume you are at least somewhat, or else wtf are you doing here), you can *easily* write software that uses, and proves, evolution.
Generate two lists of sets of random characteristics. Breed pairs of list items by selecting randomly between the characteristics contained in the items for a full set: AB, BA, AB, BA, etc. Now you have a new item with a new combination of characteristics. Assign each characteristic a weight: ability to find food, resistance to disease, etc. Create an environment that requires certain weights for survival. Test items against environment. Some will survive: return them to the list. They get to "breed" again. The others die. Each pass through the lists will vary the population in both count and characteristics.
A pass is a generation. After each generation, graph the characteristics. Guess what? That graph will rise until the fitness of all the items reaches a peak.
What you've done is created a situation where fitness is tested against stress, and higher fitness results in more survival. Subsequent generations will be more and more fit until they're all fit enough to survive.
Now add some randomness. Kill a few off just at random. These are "accidents." Make a few of the weaker ones survive anyway. These are cripples taken care of by the community, or otherwise lucky. Run the thing again. Guess what? Fitness of the population will rise again.
This is evolution in a fishbowl, and it's a very useful programming mechanism for anything where you can assign a "gene" to an approach to a problem, and "fitness" to the result of applying that approach. That's the practical side. On the fun side, you can (and I have done) write a great game where you have critters that breed, live and die using this mechanism.
Create a 2d grid. The genes are instructions, things like: "turn left if nextcell contains rock" "move forward" "turn left", "eat (fitness up)", "turn towards food" "turn away from food" "turn away from other critter" "breed" "if critter in next cell skip next gene" "if rock in next cell skip next gene" "knock heads (one critter dies)", etc. Each critter gets a list of these, randomized. Every move costs them fitness; eating gains it back. Seed the environment randomly with food and rocks and critters. Then run them by executing their genes in order. They will initially perform very poorly -- randomly. But as you breed them and the generations pass and the genes update from the highest fitness critters, you'll end up with critters that seek out food and then go breed, never running into a rock or another critter. Add animations to taste, be sure to graph fitness for the whole population, it's fascinating.
You can add complexity by adding recessive genes, more types of actions, more stuff in the environment, etc. There's really no end to what you can do. As a fun exercise, try to create a high performing critter manually, then throw it into the mix. Then at the end, when the fitness has maxed out, take a look at the highest fitness critter and see not only how little it resembles your well thought out choices, but what bizarre strategies it's implemented to be better than what you worked out. It can be mind blowing.
evolution: not only a fact, one well within your reach to test, verify without a shade of doubt, and use to your own benefit.
Once you've seen this work in practice, assuming you've got a decent head on your shoulders, you will immediately be able to generalize the process to nature and generations of real critters, from moths to humans to whatever. Strategies and capabilities against stressors, survival of the fittest, it's just the way it works, and there is ZERO doubt about it among those who actually understand it. Anyone who denies evolution is either ignorant of the facts or deliberately snowing you for some reason. 100% guaranteed. There are no other possibilitie
I've fallen off your lawn, and I can't get up.
Yes, laudable in concept. But I'd like to point out that you can't fix stupid, and trying invites heartbreak and consists of a massive waste of time. Also, you can't fix faith -- it strikingly resembles stupid in form, effect, and depth of infestation. And it's worse in one way: Being stupid is not politically correct. Exhibiting faith is. Woe is us.
I've fallen off your lawn, and I can't get up.
Wishful thinking. When Galileo's presentation was found to be at odds with religion by the ultimate arbiter of such things (the then-Pope), they certainly were in conflict. The "apology" took centuries to come around, far too late to help Galileo. It wasn't a misunderstanding. It was stubborn clinging to myth and nonsense with the added salt of the power to enforce the myth over the facts.
When religion gives us social rules, there may be, often is in fact, value to be had. When religion tries to tell us how the world came to be and why we are here, it falls flat on its face, each and every time. It's the purest form of conflict: The intentional and irresponsible promulgation of fictions in the face of repeatable, consensual facts to the contrary. The more we understand, the more visible this is.
I've fallen off your lawn, and I can't get up.
*Ignorance* is temporary. Stupid is forever.
You really have no idea how 'publish or perish' is involved?
Here's a clue: when was the last time you delayed publication (of eminently publishable results) to run some extra tests, or perform alternative forms of verification? I've never had a supervisor allow such things in my entire career. It's always a case of publishing as soon as possible (i.e. as soon as a study has the remotest chance of getting past reviewers).
I tend to be very cautious in my approach to things, and I've often wanted to do additional verification work. Not to target a better journal or a second publication, but just for the sake of more solid conclusions. I'm never allowed to do this, and I even recognise that it's not part of my job to cause any problems over it, for the very economic reasons that you mention. This bothers me deeply, but it doesn't seem to bother the kind of people who care more about their careers than about the veracity of their results. I've even been told on a few occasions that my reticence to publish some of my own simulation work that "should already be out there" is bad for my career.
In my perception, those who are more career-driven have an advantage in gaming the system. They are rewarded for publishing multiple papers of shallow scope and relatively minor significance; spreading what should be presented once as thin as possible across multiple publications. We all know it's a game to be played; that those evaluating our early-career performance really have no clue whether a publication is important or not. By the time they find out, those who've gamed the system well will already have tenure.