This week’s Nature features a call to arms from George Ellis and Joe Silk, entitled Scientific method: Defend the integrity of physics. I’m very glad to see well-known physicists highlighting the serious problem for the credibility of science raised by the string theory multiverse and the associated ongoing campaign to justify the failures of string theory by attacking the scientific method. Acknowledging evidence that an idea you cherished doesn’t work is at the core of what science is and physics now has a major problem with prominent theorists refusing to abide by this principle. Ellis and Silk do a great job of identifying and characterizing an important challenge the scientific community is facing.
The issue is however complicated, and while the Nature piece carefully and clearly addresses some of the complexities, there are places where things get over-simplified. In particular, the introduction frames the issue as whether a theory being “sufficiently elegant and explanatory” allows it to not need experimental testing. The problem with the string theory multiverse though is not this, since such a theory is the antithesis of “elegant and explanatory”. There’s just about nothing in science as inelegant as the various attempts (e.g. the KKLT mechanism) to make string theory fit with known physics, and “the multiverse did it” is no more an actual explanation of anything than “a big omnipotent turtle did it”.
Trying to cut through the complexities, Ellis and Silk write:
In our view, the issue boils down to clarifying one question: what potential observational or experimental evidence is there that would persuade you that the theory is wrong and lead you to abandoning it? If there is none, it is not a scientific theory.
This is at the heart of the matter, but there are subtleties. A common recent move among some prominent string theorists has been to argue that string theory is falsifiable: it is based on quantum mechanics, so if experiments falsify quantum mechanics, they falsify string theory. This just makes clear that the question of falsifiability can be slippery. Philosophers of science are experts at the intricacies of such questions and Ellis and Silk are right to call for help from them.
They also make the interesting call for the convening of a conference to address these issues. How such a thing would work and how it might be helpful seem well worth thinking about. As for one of their other recommendations though:
In the meantime, journal editors and publishers could assign speculative work to other research categories — such as mathematical rather than physical cosmology — according to its potential testability.
I’m leery of the impulse among physicists to solve their problem of how to deal with bad physics by calling it mathematics. Yes, there is good mathematics that has come out of untestable ideas about string theory, but no, this doesn’t include the string landscape/multiverse cop-out, which physicists need to face up to themselves.
For the specific arguments from Sean Carroll and Richard Dawid that Ellis and Silk address, I’ve written about them elsewhere, see for instance here, where I discussed in some detail Dawid’s arguments.
Update: Sabine Hossenfelder has commentary on this here.
Update: Taking the opposite side of the argument in January’s Smithsonian magazine is by colleague Brian Greene, with an article entitled Is String Theory About to Unravel?. As you might expect, Brian’s answer is “No”, and he gives a good account of the point of view Ellis and Silk are warning against. He mentions the possibility of encouraging news for string theory from the next LHC run, but says that “I now hold only modest hope that the theory will confront data during my lifetime.”
Update: Sean Carroll responds to the criticism from Ellis and Silk with a tweet characterizing them as belonging to the “falsifiability police”:
My real problem with the falsifiability police is: we don’t get to demand ahead of time what kind of theory correctly describes the world.
Update: Gordon Kane joins the fight in a comment at Nature, claiming that, before the LHC, string theory predicted a gluino mass of 1.5 TeV.
The literature contains clear and easily understood predictions published before LHC from compactified string theories that gluinos, for example, should have been too heavy to find in Run 1 but will be found in Run 2 (gluino mass of about 1.5 TeV).
As far as I can tell, this is utter nonsense, with Kane publicly claiming string theory predictions of a gluino mass of around 600 GeV (see page 22 of this) back in 2011, then moving the “prediction” up as Run 1 data falsified his earlier predictions. Kane at least makes falsifiable predictions, the problem with him only comes when they get falsified…
Update: Chad Orzel has his take here.
Update: Adam Frank has an essay on this here.