Saturday, September 24, 2011

Mr. Neutrino, I'd like you to meet Mr. Bayes

A lot of people who want to believe the faster-than-light neutrino result (and count me as one who would love it to be true!) are displaying a tendency to get quite pissy at those of us who remain skeptical, e.g. see this comment over at Bad Astronomy, which is actually one of the more mild ones. There are accusations of dogmatism, that a "true" skeptic should just follow the results.

That is true as far as it goes, but in my never-humble opinion, if you want to "follow the results" properly, you must be a Bayesian.

The prior probability, based on everything we know about physics, that neutrinos can travel faster than the speed of light in a vacuum, is vanishingly small. The results of countless past experiments would have to be either dismissed, or given a completely different explanation. Anything is possible, of course, but as somebody else said (can't remember where, sorry, no credit), this is "possible" on the order of saying, "Oops, turns out gravity doesn't attract, it repels!"

And Bayes tells us then, that even what would be very solid experimental support for this result in another context is totally unconvincing in this context. Bayes Theorem doesn't quite work this way, but to just simplify it: Let's say last week I gave you trillion-to-one odds that neutrinos could travel faster than light. Then we come up with an explanation for how this result could be an error, but unfortunately it requires a series of coincidences that are a billion-to-one against. I'd still believe it was the billion-to-one error before I'll believe FTL neutrinos were a real thing.

It's probably something much more mundane, of course. Or perhaps it will turn out to reveal some weird new physics where the neutrinos can appear to some observers to exceeding light speed, but in a way that doesn't screw up causality and relativity; that would be neat. But if you're a Bayesian (and you should be), then you can be virtually certain that it's not really FTL. Sorry, sci-fi fans. No galactic empires for you.


  1. As a time nerd(*), I'm thinking there's a completely reasonable and plausible timing-related explanation for all of this, but I haven't figured it out yet. Neither have some dozens of people who have actually put their names on a paper claiming faster-than-light transmission of neutrinos, though!

    At the very least, the results indicate that continents likely do drift.

    (*) For a good time, call 2600:3c03::13:3123

  2. On another comment at Bad Astronomy, one person speculated that one possible explanation -- and this would be really weird and would still overturn a lot of physics, but not so much that it would make the results of past experiments implausible -- is that neutrinos do not exhibit Lorentz invariance, i.e. that means they could travel faster than light without screwing up causality. I still haven't wrapped my head around whether that makes any sense at all.

    Let's hope it's at least something cool like that! And not just, "Oh, the janitor has OCD and kept resting the knobs" or something.

  3. I'm going to predict this turns out to be experimental error on pragmatic grounds. If I'm right, then I get to feel smug. If I'm wrong, I still get to be excited by all the cool new physics that will doubtless come out of this.

    Of course, xkcd got there first ;-).