When is a theory a theory? Long ago we  commented briefly on the Climategate revelations that the global warming books have been cooked to support the theory. There are a great many blogs dedicated to tracking how that miserable field is regressing, so I have felt no need to beat an obviously dead and cooling horse. But physicist blogger, Lubos Motl, questions why a 2-sigma result (1:20 chance of being accidental) of climate warming (a highly contested result, not supported by data contends Roy Spencer) should cause the American Physical Society to claim "incontrovertible proof" when a 6-sigma result  (1:Million chance of being accidental) from a neutrino detector is doubted by all concerned.

In effect,  Motl, makes the old argument that there are "hard" and "soft" disciplines within science, and that "hard" scientists tend to need more sigmas in their data, but also give much more significance to the "priors", the body of evidence that support a theory. So when a well-established theory is contradicted then "extraordinary claims require extraordinary evidence". (I'll come back to the strange origin of that phrase, and its degenerate character.) However, Lubos goes on, the "soft" sciences are so squishy, that one never gets better than a 2-sigma effect, and so the criteria are much lower by comparison. By the same token, there isn't much benefit in promoting a novel theory in the soft sciences, say, by imposing a tax on everyone, because these theories are so full of exceptions and prone to reversals. While one would hope that climate science is eventually going to be a "hard" science, the field is young and still quite "soft", which is why Motl sees the exaggerations and politics surrounding climate warming as so egregious.

My own view is that the "soft" sciences are soft for a very good reason--they involve humans. Psychology, psychiatry, sociology, political science, economics and yes, predictive climate change all involve people and populations. Since people have this annoying habit of behaving irrationally or more precisely, obstinately, (where they know the prediction and therefore do the opposite), there can never be great certainty to any process involving them. (If you are callous enough to think otherwise, then imagine that your own theorizing can be predicted accurately, then your own theory turns out to be ho-hum obvious, and you didn't need to waste our time making it anyway, much less the PhD you wrote.) The circularity of the "soft" sciences can be vicious if you pretend they are "hard".

In the case of global warming, the danger sign was including the adjective "Anthropogenic", as in AGW. Once we include humans in the mix, the theory becomes more volatile than nitroglycerine. And this is where Motl and "deniers" claim that the theory parts company with the "hard" sciences.

Even if we have to write off the "soft" sciences, at least we have recovered the objectivity of the hard sciences, or have we?

And here is where I part company with Motl. For it is true that numbers are far more compelling in the hard sciences, but it is also true that the numbers don't vote, don't write textbooks, don't teach, and don't blog. And while the theories may be quite firm and not at all circular, the same cannot be said of the making of the theories. That is, theories don't write themselves, and since theorizing is done by humans, any concerns or worries about theorizing is now just as "soft" as sociology.

Is this a problem? I mean, humans are involved in say, the making of chairs, but that doesn't prevent us from sitting on them, does it?

No, not at all. The problem isn't with the chairs, the problem is when there's a sociological upheaval and chairs get redefined. While the "hard" science remains hard and quite independent of the people who made it, the definitions could change and make it very difficult to sort out the hard from the soft. For example, at one time it was thought that seeing through walls was a supernatural gift and clearly "soft", but the discovery of x-rays redefined the field as "hard". Contrariwise, the study of "animal magnetism" was probably once a "hard" science that has since gone "soft".

So let's take a "hard" science like physics, and ask, what is the status of "string theory"? Is it "hard" or "soft"?

This is the topic of a most interesting paper that evaluates string theory with four modern philosophies of science, which is what the serious study of hard vs soft science is called. The author runs through the criteria, and concludes that the theory fails at all four, making string theory about as soft as AGW. Here's the four views and how they stack up.

Phil. of Science	What is a scientific theory?	String theory is:
Positivism	Verifiable theory	Not verifiable
Popper	Falsifiable theory	Not falsifiable
Kuhn	Incommensurate paradigm	Not even incompatible
Lakatos	Progressive vs degenerate	More likely degenerate

To recap the history of the philosophy of science, the Vienna Circle and logical positivism said a real scientific theory had to have data validating it, it had to be verifiable. Yet theories such as Marx's Communism, Freud's Psychotherapy, and even Darwin's Evolution seemed to be constantly validated without any data ever contradicting them. Thus Karl Popper parted ways with the Vienna Circle, and said that a real theory had to permit or even require that it be invalidated or falsified by some data. This led to the problem that some theories seemed to be in principle unfalsifiable, yet they otherwise functioned as fine scientific theories. Thomas Kuhn was a graduate student of a Vienna Circle member, who was given the task of defending the verification thesis by examining history. He came to the conclusion that the same data supported two opposite (incommensurate) views (paradigms), and thus the victory of one theory over another was sociological--one group of theorists died and another rose up to take their place. Imre Lakatos didn't think the various paradigms had to be incommensurate, Einstein's theory of gravity didn't deny Newton, it merely augmented it. However when a theory spends all of its time making ad hoc excuses to explain unexpected data, then the theory is in its "degenerate" phase, whereas a theory that is explaining more and more similarities with the same set of models is in its "progressive" phase. It is the greater explanatory power of "progressive" theories that makes them superior to "degenerate" theories, and how they ultimately triumph.

String theory, according to the physicists who wrote this article, is empirically degenerate but theoretically progressive. I'm not sure this distinction has any meaning. If one doesn't worry about the data, then how does one know if all these ad hoc rules are brilliant development or just epicycles? But Lakatos only suggests that one choose the more progressive program, and if there is only one program then there is nothing to switch to. The authors suggest that this is the case for string theory, there just isn't any alternative--which I find an odd position to defend. There have been many alternatives, most of which are neglected because all the money is in string theory right now.

But the process of deciding what was a "good" theory made me want to apply this to Darwinism. Here's how it shapes up:

Phil. of Science	What is a scientific theory?	Darwinism is:
Positivism	Verifiable theory	Not verifiable
Popper	Falsifiable theory	Not falsifiable
Kuhn	Incommensurate paradigm	Not even incompatible
Lakatos	Progressive vs degenerate	More likely degenerate

Strangely enough, all the same arguments apply to Darwinism that applied to string theory! Like string theory, it isn't verifiable, or even falsifiable. It is not even incommensurate with any other theory, because it tries to be everything to everyone. When difficulties arise, say altruism or convergent evolution, it makes up rules to handle all the exceptions. So it operates as if it is in degenerate mode. And people are fond of saying that despite its faults, there really is no alternative to the theory that might be more progressive.

So if this be the mode of the 21st century, science will be inundated in "grand unified theories" that suck up all the resources without making much progress--hindering the development of real alternatives. Which brings us to the phrase "extraordinary claims require extraordinary evidence". To my knowledge, the statement was first made in 1976 by Carl Sagan in his own defense. He was a co-investigator on the mass spectrometer experiment on Viking that failed to find life on Mars. His civil engineering colleague, with a spectacularly simple instrument (the Labelled Release) did find life. Since it was informally thought that the discovery of life would be an automatic Nobel Prize, Sagan had to explain why his PhD-loaded team didn't find it but a civil engineer did. So Sagan intoned this famous line, but of course he reserved the right to decide whose data was the more extraordinary. And no surprise, the status quo was upheld. This is what Lakatos called "the protective belt" of ad hoc statements to protect the theory. It is a sign of a degenerate science program.

Kuhn may not have been surprised, but Lakatos surely would have been disappointed.