Tuesday, February 19, 2013

Rules of thumb and science announcements

Texan99 has a question over at Grim's Hall about how a layman can evaluate claims by scientists. It isn't easy to come up with a simple rule, and I thought maybe I should elaborate on my comments there.

Pronouncements in the press tend to be pretty dramatic. Drama is pretty much the only way to get into the news, so sometimes it is the scientist and sometimes the reporter who gooses the story.

Always remember that what you read is typically what comes out after a reporter has digested the information.

Pronouncements fall into several categories:

  1. Solidly backed by experiment and theory. Sometimes reality is weird (e.g. quantum mechanics) and you need some experience to understand it; and reality is often inconvenient. Look for humility, stable consensus and eagerness to talk about experimental verification, but those aren't perfect guarantees.
  2. Within the bounds of a current consensus that works OK. The consensus might be wrong, but that would be a big deal. Sometimes we know the current theory has holes, but nothing has been better so far. For example, quantum mechanics and theories of gravity don't play well together, so string theory (which is supposed to unify them) has been popular for a couple of decades, despite the fact that it hasn't gotten anywhere. The rest of us muddle along with two inconsistent theories and apply each in the place where it works best.
  3. Badly scrambled: either this is speculative and not really in the consensus or the reporter garbled it. This, for example. The theories about universe bubbles are part of some cosmological theories, but they're quite speculative, they're untestable, and connections to particle physics are even more speculative. The reporter was hunting for something weird enough to print, and I doubt that Hill or Lykken was happy with the story. Look for something that would be hard to test, or which doesn't seem tightly related to the experiment at hand.
  4. Within a current consensus that doesn't work worth beans though we pretend it does (some fads in psychiatry come to mind). AGW is so political that it is hard to have a science discussion about it, although you may have noticed that the "A" (anthropogenic) and "W" (warming) have pretty much disappeared from the media reports. The consensus didn't work so well... If you don't have domain knowledge of the field, about the best you can do is know a little of its history. If the consensus changes frequently, don't trust it. (Don't trust the pronouncements or the consensus.) But the only way to know the history of a field is to either study it (which isn't easy to do) or live long enough and pay attention. Watch for supporters with vested interests.
  5. A wondrous new paradigm shift with new science and new vistas. This is very very rare. I can think of "jumping genes," but not too much else recently. Not something you need to worry about most of the time.
  6. Botched experiment. This is actually rather common, as scientists eager for publication and possible future research grants rush out press releases before the peer review. Hint: is the news about something that has been accepted for publication? If not, be cautious; and even if it has been accepted, know that there are some vanity science publications out there...
  7. Hollow. We've no shortage of "theories" about healing energies or ways to disprove Einstein or allow perpetual motion. In the physical sciences these are usually easy to spot if you have the habit of trying to be precise about meanings of words or of looking for what a theory predicts mathematically. I'm not so sure how to spot them in linguistics or neurology or psychology. A rule of thumb: if the speaker explains how he is overthrowing current paradigms or going against an intrenched establishment, he's usually full of it. But not always. Sometimes the intrenched establishment is full of it (see above), and sometimes they both are.
  8. "What a noble mind is here o'erthrown." Doing important work in one field doesn't mean you're going to be right at whatever your hand finds to do. Linus Pauling was a genius, but got a bee in his bonnet about vitamin C and lost track of his own disciplines. Look for simple solutions to hard problems; chances are they're too simple and the visiting expert has left some things out.

Just a few possibilities ... You can also win a little by looking for poison citations. Someone who cites Fort, or has research that cites only his own work (Noam Chomsky)--dubious. Each field has its own usual suspects. Perpetual motion or disproving special relativity are obvious ones in physics, but I don't know what their counterparts are in archaeology or embryology. Maybe somebody should make a list.

7 comments:

Texan99 said...

"If the consensus changes frequently, don't trust it." That seems excellent advice, especially if you're being rush to make up your mind to give someone a bunch of money or power to fix the problem, because waiting until it's clearer will be fatal.

Re (7), I have two words: cold fusion.

Who is "Fort"?

Maybe I just haven't been paying enough attention in recent decades, but the whole AGW thing is very discouraging to me. I trusted the process to do better than that. Assuming there is such a thing as a truly professional climate scientist, this is the first time I've ever found myself in the position of trying to contradict scientists in their own fields, when a lot of them really seemed to agree. It's been messing with my head for years now. Is this CO2 forcing business really just so much phlogiston, or am I in over my head?

james said...

Charles Fort was a professional contrarian.

"Cold fusion" is a very odd duck. Pons and Fleischmannn I think I can safely discount as sloppy and dishonest. However, you can expect fusion to happen at a slow rate, and Jones (who later became a 9-11 truther) made a simultaneous measurement that didn't seem crazy.
The idea is that if hydrogen nuclei can be brought close enough together their wave function overlap will result in fusion at a small but measurable rate. One such way to bring them close is is to dissolve hydrogen (deuterium IIRC) in platinum or one of the similar metals. The lattice spacing is much closer than the spacing between hydrogen atoms in a molecule, even at tremendous pressure. If you can make the lattice even tighter yet you could make the reaction go faster. If you heat up the metal some of the vibrations will bring the hydrogen atoms closer together for short periods of time.

The funny thing is that some attempts to do this seem to have had mildly positive results, and others negative results. Some of the "successful" experiments used high temperature and pressure; but unfortunately that makes delicate measurements hard. Too much massive hardware is in the way.

Maybe there is a phase transition in the platinum crystal lattice. That might explain the varying results. (But not P&F's--I judge that they goobered it up and then wouldn't admit it.)

The AGW debacle is very discouraging. It seems to arise from two problems. One is that when a field is broad enough and the specialty narrow enough, the circle of peers for peer review can be very small even though there are lots of researchers in the field. A failure can go uncaught for quite a while that way, as we saw with Bateman's work with fruit flies.

The other problem comes when greed (more grants, more fame!(*)) meets with political interests which meets with media bias and makes a perfect storm. If we're destroying the world (a popular media theme) we need to be taken in hand by our betters (political hacks) and guided in the way we should go (another popular media attitude).

It doesn't help that the effect they're looking for is rather subtle and hard to precisely define. An overall increase in temperature will make local temperatures rise or fall: you have natural variations you have to account for and then you have to figure out a way of weighting the temperatures measured in different places in the world. If you had thermometers everywhere you could just average every reading in them all for a year, and compare it with previous years to see if the average was rising or falling. But we can only use the thermometers we have, which are very far from uniformly spaced.(**)

james said...


I'd hate to try to do that kind of weighting; there's too much possibility of inadvertent bias, and there are no good models you could use to tune your analysis on before unblinding.

That there can be global warming and cooling isn't particularly controversial. Measuring it is tricky, and so is properly attributing its cause.

I used to be willing to take their results at face value, including the old hockey stick, even though the hockey stick graph shows a characteristic pathology of a failed fit. When I learned that it depended on a single tree-ring sample set, that was the end. The leaked emails weren't as much of a smoking gun as advertised--the attempt at intimidation was very bad but the "massaging the data" wasn't obviously bogus. But their data presentation was sometimes deceptive (presenting corrected data as though it were the original!) and it sounds as though they lost so much of their metadata over the years that nobody will ever be able to try to repeat the results.



(*)Scientists are very often motivated by desire for fame.

(**)And does a 2 degree rise in dry air temperature match a 2 degree fall in damp air somewhere else? You have to take humidity into account too, I guess.

Texan99 said...

Was the platinum cold-fusion thing the one from the early 90s? I wasn't qualified to understand it, but my father was living at that time, and he was. If I'm thinking of the right one, it was his impression that they had constructed a very expensive battery that consumed platinum.

The problem with the hockey stick, in part, was that its dramatic inflection point happened just at the point where they switched from one kind of data to another. Something analogous to that plagues the new models: they have to assume that the CO2 feedback mechanism of the last 125 years is changing to a heightened sensitivity more or less at the present moment. So the curve is discontinuous right at the present -- meaning that there's no observational data yet to support the discontinuity.

james said...

Yes, the platinum was the one. I always suspected that it was leaking oxygen into the mix and catalyzing the hydrogen, but nobody was allowed to study their rig.

You can get similar sorts of fit problems when the bad points are other places, not just the inflection points. Makes it hard to figure out where the problem lies...

Sam L. said...

And the end product of digestion is...

#4--all supporters have vested interests. Skeptics, too.

Cold fusion--I keep seeing the occasional mention in ANALOG science discussions that there seems ("") to be something happening. Over the last 5 years or so.

james, we're always "destroying the world".

Yannow, there's massaging the data, and there's beating the hell out of it, and both occur in AlGorebal Worming.

james said...

I read a literature review on cold fusion about 5 years ago. The results were split, but it was clear that P&F's results were not reproducible. And _if_ something was happening, there wasn't a good model of exactly what.

The CRU crew were also presenting massaged data as though it were the reals measurements without any hint that it wasn't the raw data.