Michael Douglas gave a colloquium at City College this afternoon, with the title “Are there testable predictions of string theory?” I went up there to the talk, figuring that I knew more or less what he would say, but he really surprised me. Douglas has given many talks over the last year or so about his program for trying to get predictions out of string theory by doing statistical analyses of string vacuum states. He has concentrated on what looks like the most promising case, trying to see whether vacua with low-energy supersymmetry breaking are favored over ones where supersymmetry is broken at much higher scales (e.g. the GUT or Planck scales). If he could make a prediction that the LHC will see supersymmetry, that would count as the first real prediction of string theory, and in 2008 or so we would see if it was right. I was expecting Douglas today to explain this whole program, report on what he had achieved so far, and offer hope that he and his collaborators would have a yes or no answer about supersymmetry sometime soon.
Instead he very much downplayed hopes for this kind of prediction, answering a question about it by Nair at the end of the talk by explaining some of the difficulties. Presumably he now agrees with recent claims by Dine that it is too difficult, even in principle, to decide whether or not the landscape predicts supersymmetry. Given this, in the conclusion of his talk, I was expecting him to answer the “Are there testable predictions” question in the negative. Instead, he did something very strange. He announced that string theory does make predictions, lots of them, adopting the Lubos Motl definition of a “prediction” of string theory as being anything consistent with string theory. Examples he gave included Polchinski’s cosmic string networks, where one could tell from the behavior of the network whether the strings were fundamental or not, and short distance modifications to GR. Of course these are not in any sense real predictions; all sorts of different modifications of GR at short distances are compatible with string theory, as are either no visible fundamental cosmic strings, or visible ones with a huge variety of possible different properties.
The weirdest part of his talk was when he explained what he considered the best prediction of string theory. This involved the negative prediction that the fine structure constant can’t have varied with time in the early universe, since effective field theory arguments would imply a corresponding variation in the vacuum energy, something inconsistent with observation. So his best prediction from string theory isn’t really a prediction of string theory at all, but actually a prediction of effective field theory. Furthermore this “prediction” is the purely negative one that something that hardly anyone expects to be true actually isn’t true.
In the question section, some obnoxious guy who has a weblog asked him whether it was really true that the best prediction string theory could come up with was the no variation of the fine structure constant one that was really an effective field theory prediction, and didn’t that mean there was no hope of string theory ever really predicting anything. For some reason this made him rather defensive, and he began by saying it depended on the meaning of the word “prediction”. After having it explained to him what most physicists consider a prediction to be, he launched into a sequence of analogies designed to explain why you can’t get real predictions out of string theory. They all were of the same genre: imagine some situation where you can only observe phenomena that are related in a very complicated and hard to calculate way to the underlying fundamental theory, and somebody tells you what the fundamental theory is. Shouldn’t you work on it and believe in it?
This argument makes it clear where the whole subject is going to end up. The standard scientific method of deciding whether a theory is true or not by figuring out its implications and comparing them to observations is no longer operative. In the case of string theory there’s a new method. You just believe because authorities tell you to, and from now on the activity of professional theorists will consist solely in the construction of elaborate scenarios designed to explain why you can’t ever predict anything. Feynman’s line that: “string theorists don’t make predictions, they make excuses” has been changed from a criticism into a new motto about how to do science.