I recently noticed that, around the same time I was preparing my slides for a talk about Is String Theory Testable?, Michael Douglas was doing something similar, preparing a talk on Are There Testable Predictions of String Theory? There’s a certain amount of overlap in our presentations, and people might find it interesting to compare them.
Douglas goes over much the same story I do, but reaches different conclusions. For him, string theory does “make predictions”, just lots and lots of incompatible ones, so the problem is that:
none of the ideas which have been suggested so far are guaranteed signatures of string theory. We would be happier with one prediction, which could lead to a decisive answer either way.
This is the sort of thing I would call a prediction, so I guess we agree that they don’t exist. Douglas ends by noting that the one way he can think of to get such a prediction is through a statistical argument based on counting vacua and the dynamics of eternal inflation. He doesn’t mention the argument given in my talk that this is already falsified (by the lifetime of the proton), or that such arguments inherently lead to calculations that are inherently intractable and can never be done (this argument is due to him and Denef, it is surprising that he doesn’t mention it).
Along the way, Douglas does make a couple claims about things that he thinks the statistical anthropic landscape arguments disfavor, especially varying constants and large distance modifications to gravity. Seeing these would falsify our current theory, but would not falsify string theory, since it can accomodate them, although perhaps not within Douglas’s statistical framework.
One of the more surprising responses I’ve seen to my recent claims that string theory has failed as an idea about unification because it’s inherently untestable comes from Mark Srednicki, who writes (in the context of mentioning an MSNBC interview with Brian Greene):
We see that the big issue for Brian, and for just about all scientists (though with the apparent exceptions of Lee Smolin and Peter Woit), is what is TRUE. Not what corresponds to some philosophy of what science is or is not. Lee writes that the landscape must be rejected because “it would mean the end of our field” (page 165). It should be obvious that this is not the basis that is traditionally used for accepting or rejecting a theory! Peter’s (essentially the same) argument that string theory must be rejected because (at the moment) it does not appear to be sufficiently predictive (for Peter) is also irrelevant to the question of whether or not string theory is TRUE.
If the landscape is right, we may never get anything more than circumstantial evidence that it’s right. But that’s often the case in science. We’ve been spoiled in particle physics by having extremely precise data and highly predictive and quantitative theories for the past few decades. Most of the rest of science has not been so lucky. Perhaps we will not be so lucky going forward. The only way to find out is to do more work and see where it leads.
Srednicki’s reaction to the lack of testability of string theory seems to be that testability is not what matters. What matters is what is TRUE, and it’s perfectly logically consistent for string theory to be of such a nature that we can never test it. The problem with this point of view is that science is not so much about what is true as about how one knows what is true about the world. Religious believers are also interested in what is true and think they know what this is, but science is different since it provides a means for deciding what is true. Scientific ideas about the universe are true when they make predictions that can be checked in a convincing way. Ideas that can’t be experimentally checked in some way, directly or indirectly, may or may not be true, but they’re not scientific ideas, rather something of a different nature.