Thursday, January 18, 2018

Multiverse and falsifiability: W*it vs Carroll

Sean Carroll has released an essay elaborating upon his talk at Why Trust a Theory, a late 2015 meeting organized by TRF guest blogger Richard Dawid:
Beyond Falsifiability: Normal Science in a Multiverse (arXiv, Jan 2018)
On his self-serving blog, Carroll promoted his own preprint.

Well, once I streamline them, his claims are straightforward. Even though we can't see outside the cosmic horizon – beyond the observable Universe – all the grand physical or cosmological theories still unavoidably have something to say about those invisible realms. These statements are scientifically interesting and they're believed to be more correct if the corresponding theories make correct predictions, are simpler or prettier explanations of the existing facts, and so on.

There's a clear risk that my endorsement could be misunderstood. Well, I think that Sean Carroll's actual papers about the Universe belong among the bad ones. So while I say it's right and legitimate to be intrigued by all these questions, propose potential answers, and claim that some evidence has strengthened some answers and weakened others, it doesn't mean that I actually like the way how Carroll is using this freedom.

In particular, his papers that depend on his completely wrong understanding of the probability calculus – and that promote as ludicrously wrong concepts as the Boltzmann brains – are rather atrocious as argued in dozens of TRF blog posts.

Peter W*it isn't quite hysterical but unsurprisingly, he still trashes Carroll's papers. According to W*it, Carroll is arguing against a straw man – the "naive Popperazism" – while he ignores the actual criticism which is that the evaluation of these multiverse and related theories (and even all of string theory, according to W*it) isn't just hard: it's impossible. The straw man is the claim that "things that can't be observed should never be discussed in science". W*it asserts that he has never made this claim; well, I would disagree because that's what he has said a few times and what he wanted his fans to believe very many times.

But let's ignore that the straw man isn't quite a straw man. Let's discuss W*it's claim that it's impossible to validate the multiverse-like cosmological theories even in principle. Is it impossible?

Well, it just isn't impossible. The literature is full of – correct and wrong – arguments saying that one theory or one model is more likely or less likely because it implies or follows from some results or assumptions that are rather empirically successful or unsuccessful. I found it necessary to say that the literature sometimes contains wrong claims of this type as well. But they're wrong claims of the right type. The authors are still trying to do science properly – and many other scientists do it properly and it's clearly possible to do it properly, even in the presence of the multiverse.

As Carroll correctly says, all this work still derives the scientific authority from abduction, Bayesian inference, and empirical success. For example, Jack Sarfatti has a great scientific authority because he was abducted by the extraterrestrial aliens. ;-) OK, no, this isn't the "abduction" that Carroll talks about. Carroll recommends "abduction" as a buzzword to describe the "scientific inference leading the scientists to the best explanation". So "abduction" is really a special kind of inference or induction combined with some other methods and considerations that are common in theoretical physics and a longer explanation may be needed – and there would surely be disagreements about details.

If you're a physics student who knows how to do physics properly, you don't need to know whether someone calls it inference, induction, or abduction!

But it's possible to do science even in the presence of unobservable objects and realms that are needed by the theory. The theory still deals with observable quantities as well. And if the agreement with the observed properties of the observable entities logically implies the need for some particular unobserved entities and their properties, well, then the latter are experimentally supported as well – although they are unobservable directly, they're indirectly supported because they seem necessary for the right explanation of the things that are observable.

Also, W*it observes that "some theoretical papers predict CMB patterns, others don't". But even if one proposes a new class of theories or a paradigm that makes no specific observable CMB or other predictions, it may still be a well-defined, new, clever class of theories and the particular models constructed within this new paradigm will make such CMB predictions. Because the discovery of the class or the paradigm or the idea is a necessary condition for finding the new models that make (new and perhaps more accurate) testable predictions, it's clearly a vital part of science as well – despite the particular papers' making no observable predictions!

Peter W*it has never done real science in his life so he can't even imagine that this indirect reasoning and "abduction" – activities that most of the deep enough theoretical physics papers were always all about – is possible at all. He's just a stupid, hostile layman leading an army of similar mediocre bitter jihadists in their holy war against modern physics.

There's another aspect of W*it's criticism I want to mention. At some moment, he addresses Carroll's "another definition of science":
Carroll: The best reason for classifying the multiverse as a straightforwardly scientific theory is that we don’t have any choice. This is the case for any hypothesis that satisfies two criteria:
  • It might be true.
  • Whether or not it is true affects how we understand what we observe.
Well, I am not 100% certain it's right to say that we can't avoid the multiverse. On the other hand, I understand the case for the multiverse and I surely agree that physics is full of situations in which "we don't have a choice" is the right conclusion.

Peter W*it doesn't like Carroll's quote above because it also allows "supreme beings" as a part of science. I don't see what those exact sentences have to do with "supreme beings" – why would an honest person who isn't a demagogue suddenly start to talk about "supreme beings" in this context. Nevertheless, I see a clear difference in W*it's recipe what science should look like and it's the following principle:
Whatever even remotely smells of "supreme beings", "God", or any other concept that has been labeled blasphemous ;-) by W*it has to be banned in science.
W*it hasn't articulated this principle clearly – because he doesn't have the skills to articulate any ideas clearly. But one may prove that he has actually applied this principle thousands of times. Apologies but this principle is incompatible with honest science that deals with deep questions.

Important discoveries in theoretical physics may totally contradict and be the "opposite" of stories about "supreme beings" (or any other "unpopular" concepts); but they may also resemble the stories about "supreme beings" in any way – in a way that simply cannot be "constrained" by any pre-existing assumptions. The correct theories of physics must really be allowed to be anything. Any idea, however compatible with Hitler's or Stalin's or W*it's ideology or political interests, must be allowed to "run" as a candidate for the laws of physics.

W*it clearly denies this basic point – that science is an open arena without taboos where all proposed ideas must compete fairly. He wants science to be just a servant that rationalizes answers that were predetermined by subpar pseudointellectuals such as himself and their not terribly intelligent prejudices.

That's not what real good science looks like. In real good science, answers are only determined once a spectrum of hypotheses is proposed, they are compared, and one of them succeeds in the empirical and logical tests much more impressively than others. Only when that's done, the big statements about "what properties the laws of physics should have" can be made authoritatively. W*it is making them from the beginning, before he actually does or learn any science, and that's not how a scientist may approach these matters.

If the vertices in the Feynman diagram were found to be generalized prayers to a supreme being, and if the corresponding scattering amplitudes could be interpreted as responses of the supreme being that generalize God's response to Christian prayers ;-), then it's something that physicists would have to seriously study. I don't really propose such a scenario seriously and my example is meant to be a satirical exaggeration (well, even if such an analogy were possible, I still think it would also be possible to ignore it and avoid the Christian jargon completely). But I am absolutely serious about the spirit. Whether something sounds unacceptable or ludicrous to people with lots of prejudices should never be used as an argument against a scientific model, theory, framework, or paradigm. (See Milton Friedman's F-twist for a strengthened version of that claim.)

That's why the influence of subpar pseudointellectuals such as W*it on science must remain strictly at zero if science is supposed to remain scientific – and avoid the deterioration into a slut whose task is to improve the image of a pre-selected master ideology or philosophy in the eyes of the laymen. Just like it was wrong for the Catholic church to demand that science serves the church or its God, it was also wrong to demand science to serve the Third Reich or the Aryan race or the communist regime, and it is wrong to demand that science must serve the fanatical atheists or West's leftists in general.

P.S.: In a comment, W*it wrote:
Rod Deyo,
Polchinski provided a reductio ad absurdum argument against the Bayesianism business in a paper for the same proceedings as the Carroll one. He calculated a Bayesian probability of “over 99.7%” for string theory, and 94% for the multiverse.
99.5% of this stuff written by W*it is composed of bullšit because Joe Polchinski was 97% serious.

Polchinski realizes that none of these values is "canonical" or "independent of various choices" and he likes to say (and explicitly wrote in his explanation of his usage of the Bayesian inference) that the Bayesian reasoning isn't the main point – physics is the main point – but he simply wanted to be quantitative about his beliefs and these are the actual fair subjective probabilities for the propositions he ended up with. That's completely analogous to the number 10% by which Polchinski once quantified his subjective belief that a cosmic string would be experimentally observed in some period of time (I forgot whether it was "ever" or "before some deadline"). I have repeatedly written similar numbers expressing my beliefs. It makes some sense. We don't need to talk in this way but we may and it's sometimes useful.

So Polchinski hasn't provided any argument against the Bayesian inference. He has pretty much seriously used the Bayesian inference in a somewhat unusual setup.

No comments:

Post a Comment