[bonsaitree]

Summary: Imperfect system is imperfect.

'Ya gotta love a tautology adorned as journalism.

Are there any reasonable alternatives for privately funded research? No. There will always be the urge to game data to fit the hypothesis and honest people will still make mistakes. This forever shall remain a challenge for science.

I'm struggling with the urge to flag, but recognize the applicability to HN.

[jerf]

I fail to see your point. The core claim that there are people who put far too much stock in the words-treated-as-magic "peer review" is definitely true; I encounter them all the time, to the point that I've also come to recognize it as a pattern. The fact that this is not true is, as you say, tautological. But some people need this fact brought to their attention, as it is clearly not generally understood.

I've come to classify attitudes about science now into three broad categories. Hostilty: "I don't care what science says, Homeopathy worked for me." Religious: "Peer review is never wrong, and therefore the consensus is also always correct in every particular and those who disagree are heretics of science!" Realistic: "Science is the best way we have to find truth, but it isn't perfect either. The history of science should be learned and carefully examined, and it should not be forgotten that everything we see in the past is still around today, the only question is where." Rather a lot of people fit into that second one, and think it is the best understanding of science there is, and tend to label people with the third level of understanding as heretics on par with homeopaths, too. Trying to prod people out of religion-of-science into realistic-understanding-of-science is a noble goal, worthy of a blog post.

Science has been wrong before, for decades at a time in every discipline I've ever studied enough to learn about its history (in addition to the current consensus), and the question is not whether there are disciplines every bit as wrong right now, but which ones they are. If you don't understand that or haven't internalized it, you don't understand the process of science, even if you can recite its tenants at will. When we say that science is a process, not an end-result, that isn't just words, it is the reality, and what it means in practice is that the consensus is very frequently wrong. There's nothing to "correct" in consensus if it's presumed always correct. The religious idea of science makes no sense; consensus must be wrong at some points in time, it is a necessary step.

(This is all discipline-neutral, by the way. I do have my opinions about which disciplines are most likely to be wrong right now, but I freely acknowledge those are opinions. Certainly in the past entire disciplines have made glaringly-obvious-in-hindsight-and-common-sense mistakes before and it is no great insight to think that there might still be such errors around. But again, I freely admit that only time will really tell, and there are certainly other wrong-consensus-views I agree with.)

[jseliger]

Great post. A lot of what you're describing is implied in Richard Feynman's "Cargo Cult Science:" http://www.lhup.edu/~DSIMANEK/cargocul.htm , which is also available in Surely You're Joking, Mr. Feynman!. The whole book is recommended.

So is Thomas Kuhn's The Road Since Structure, which collects many of his papers on, mostly, the philosophy and history of science, which are more interesting than I would've thought.

[dantheman]

I also think a big part is how forthright a scientist is. Most scientists claim way more than can be supported by their evidence, and it's not always clear what assumptions are being made. There's a great talk, I can try to find it if there is interest, that discuss the hubris that most climate scientists have. Observing a system for a tiny portion of a cycle can in no way tell you the entire cycle. If you then admit that "hey we don't know, but here's our best guess, and here's why we think it" then you've got something. Once something is established like like copper conducts electricity or the speed of light then you can say "hey we're pretty sure about this". It has a lot to do with the maturity of the science and the tools and techniques available to the scientists. Complex systems that cannot be decomposed and upon which experiments can not be made have should have a much lower confidence value than repeatable tried and true experiments. To conflate them is to cause confusion and to weaken the trust in more established science. Peer review is more about how to position your work in relation to others for ease of understanding than to verify the results independently.

[thaumaturgy]

> Most scientists claim way more than can be supported by their evidence...

Ironically, you've made a claim here which is unsupported by any evidence at all. ;-)

Seriously though, the thing about this is that research and experimental scientists are working in environments that you and I know very little about. Their claims are an extension of their experience in that environment; they spend a great deal of time for example designing experiments or attempting to reconcile large volumes of data, and when they've spent months -- or, more likely, years -- doing that, they might publish a paper.

Then the public occasionally gets ahold of this brief summary of the scientist's work over that period, and they object that the claims are specious or sparse or incomplete.

This is not to say that a scientist's claims should be trusted just because they're a scientist, but rather that the only acceptable challenge to a shortcoming in a scientific presentation is ... more science, not argument or conjecture.

To put it another way: it is extremely unlikely that a layman will find a consequential error in a peer-reviewed scientific paper published in a reasonable journal, especially without any domain knowledge in that particular field.

[dantheman]

Errors in science are well known and are part of the process, I didn't think I needed to provide evidence of that. Unfortunately science that deals with complex systems is much more difficult than those that don't. For instance, just look at nutrition research, look at the history of the food pyramid and you see poorly understood science having a substantial impact on those who are not experts because the nutrition experts were scientists, and physics works well so nutrition must also be well understood.

I just want people to bend over backwards to tell me the ways in which they might be wrong, this is useful because it helps you understand the limits of understanding and can allow people to make informed decisions.

[thaumaturgy]

I wasn't being hostile (this time), I was just having a little fun.

What you're grappling with here is basically the "I don't know what I don't know" problem. Other people on HN have talked about this before, but basically, there's the knowledge you know, the knowledge you know you don't know, and the knowledge that you don't know you don't know. How can you possibly begin to challenge an assertion in a scientific paper if you don't even know where the mistakes might be?

(I'm going through something similar right now in a criminal case. I have no background in law. I do not trust the attorneys involved. It's maddening.)

I agree that it would be nice if as a matter of habit there were a brief statement in papers and presentations outlining incomplete areas of the research or areas justifying further study. However, I would not want that to evolve into an onus on the part of the working scientists to educate every layman with a passing interest in their research.

If you really want to challenge something in a specific field, the best thing you can do is immerse yourself in that field for as long as it takes to develop an understanding of the basic principles. Demanding much else is merely intellectual laziness.

[dantheman]

I didn't think you were being hostile, and I hope you don't detect any hostility from me, none is intended.

There are many incentives to overstate the validness of ones claims and unfortunately it weakens the overall strength of the field. I'm not asking that they make it understandable to the layman, but perhaps on any recommendation they could put confidence intervals on predictions.

Here's a presentation by Carl Wunsch, and here are the slides: http://web.mit.edu/esi/symposia/symposium-2009/2009-symposiu...

http://cdn.static.viddler.com/flash/simple_publisher.swf?key...

He's one of the worlds experts on the ocean, and is saying the same thing i am.

Here's a feynman quote on it:

I would like to add something that's not essential to the science, but something I kind of believe, which is that you should not fool the layman when you're talking as a scientist. I am not trying to tell you what to do about cheating on your wife, or fooling your girlfriend, or something like that, when you're not trying to be a scientist, but just trying to be an ordinary human being. We'll leave those problems up to you and your rabbi. I'm talking about a specific, extra type of integrity that is not lying, but bending over backwards to show how you are maybe wrong, that you ought to have when acting as a scientist. And this is our responsibility as scientists, certainly to other scientists, and I think to laymen. -- Richard Feynman, CARGO CULT SCIENCE

Also to be clear, I'm optimistic and hope we can solve all sorts of problems and discover how the world works -- but I think to we need to have integrity and maintain to foster and maintain that trust. The medical field is ripe with wild extrapolation from small studies, it's a sign of exhuberance -- "Hey, here's the answer to the problem you having!" When it should be, "It seems that in some situations this may help, we still have to do more studies and don't know why it works but it seems to." It's more honest and when things go wrong, which they almost always do the person understood that it wasn't well understood.

I encourage you to watch the video - it's by an expert saying that we don't have enough information to be sure about anything and he proposes the following advice:

In the meantime, study the system; study the control options; but----take precautions (mitigation), prepare for adaptation, and do nothing that is even possibly irreversible (which may be almost anything, including of course, the ongoing injection of greenhouse gases).

Also his example of gravity waves, I think is extremely enlightening.

[thaumaturgy]

Nicely said.

For me, there are two caveats to this: 1) the mere possibility that a discipline of science might be wrong does not mean it is wrong, and unfortunately the possibility of wrongness is often used as an argument where it shouldn't be; 2) the various sciences tend to be wrong in a very specific way, and for some reason this often gets ignored in discussions like this.

The last time I can think of the popularly-accepted science of a discipline being "wrong" in exactly the opposite way from what later came to be accepted as "right" was the gradual shift from a terra-centric view of astronomy to a relativistic one. It turned out that not only was the Earth not the center of the universe, but the universe doesn't even have a center.

Since then, science -- and mathematics -- have been wrong in the sense that they are continually being revised, not reversed. Relativity represents a revision of Newton's laws, not a reversal, etc.

I think that there is less and less -- though not non-zero -- opportunity for any discipline of science to discover that its foundations are completely incorrect.

[jerf]

Reversals occur too. Tectonic plates, stress causes ulcers, neural networks are the one true future of AI, and those are just three off the top of my head. Rare for an entire science to be rewritten, sure, but trying to contain it by saying it's always just refinements betrays an unfamiliarity with scientific history too.

[CWuestefeld]

in practice is that the consensus is very frequently wrong

I'd make that statement even stronger: the consensus is almost certainly wrong. The variables are just in what manner it's wrong, and to what degree.

It's not all a revolutionary shift to a Copernican cosmos. Most things we know pretty well today, but ought to fully expect to have tweaked tomorrow. The cliche example is of Newtonian dynamics being improved by Einstein, which is still being refined through quantum mechanics, etc.

[InclinedPlane]

That's not the point. It's known that the system is imperfect.

The problem is that some people, especially those outside academia/research, have a different idea of what "passed peer review" means and a different idea of how the machinery of science works. Passing peer review doesn't mean that something has been proven as fact, it merely means (one hopes) that the authors did their due dilligence and put together research that seems solid. Put another way, it means that the reviewers don't think that the paper is obviously BS.

Peer review is just a spam filter.

The danger comes when people elevate passing through the spam filter with scientific proof.

Peer review is the start of the debate, not the end of it.