Wednesday, December 21, 2016

Reasoning in Physics

I’m just back from a workshop about “Reasoning in Physics” at the Center for Advanced Studies in Munich. I went because it seemed a good idea to improve my reasoning, but as I sat there, something entirely different was on my mind: How did I get there? How did I, with my avowed dislike of all things -ism and -ology, end up in a room full of philosophers, people who weren’t discussing physics, but the philosophical underpinning of physicists’ arguments. Or, as it were, the absence of such underpinnings.

The straight-forward answer is that they invited me, or invited me back, I should say, since this was my third time visiting the Munich philosophers. Indeed, they invited me to stay some longer for a collaborative project, but I’ve successfully blamed the kids for my inability to reply with either yes or no.

So I sat there, in one of these awkwardly quiet rooms where everyone will hear your stomach gargle, trying to will my stomach not to gargle and instead listen to the first talk. It was Jeremy Butterfield, speaking about a paper which I commented on here. Butterfield has been praised to me as one of the four good physics philosophers, but I’d never met him. The praise was deserved – he turned out to be very insightful and, dare I say, reasonable.

The talks of the first day focused on multiple multiverse measures (meta meta), inflation (still eternal), Bayesian inference (a priori plausible), anthropic reasoning (as observed), and arguments from mediocrity and typicality which were typically mediocre. Among other things, I noticed with consternation that the doomsday argument is still being discussed in certain circles. This consterns me because, as I explained a decade ago, it’s an unsound abuse of probability calculus. You can’t randomly distribute events that are causally related. It’s mathematical nonsense, end of story. But it’s hard to kill a story if people have fun discussing it. Should “constern” be a verb? Discuss.

In a talk by Mathias Frisch I learned of a claim by Huw Price that time-symmetry in quantum mechanics implies retro-causality. It seems the kind of thing that I should have known about but didn’t, so I put the paper on the reading list and hope that next week I’ll have read it last year.

The next day started with two talks about analogue systems of which I missed one because I went running in the morning without my glasses and, well, you know what they say about women and their orientation skills. But since analogue gravity is a topic I’ve been working on for a couple of years now, I’ve had some time to collect thoughts about it.

Analogue systems are physical systems whose observables can, in a mathematically precise way, be mapped to – usually very different – observables of another system. The best known example is sound-waves in certain kinds of fluids which behave exactly like light does in the vicinity of a black hole. The philosophers presented a logical scheme to transfer knowledge gained from observational test of one system to the other system. But to me analogue systems are much more than a new way to test hypotheses. They’re fundamentally redefining what physicists mean by doing science.

Presently we develop a theory, express it in mathematical language, and compare the theory’s predictions with data. But if you can directly test whether observations on one system correctly correspond to that of another, why bother with a theory that predicts either? All you need is the map between the systems. This isn’t a speculation – it’s what physicists already do with quantum simulations: They specifically design one system to learn how another, entirely different system, will behave. This is usually done to circumvent mathematically intractable problems, but in extrapolation it might just make theories and theorists superfluous.

It then followed a very interesting talk by Peter Mattig, who reported from the DFG research program “Epistemology of the LHC.” They have, now for the 3rd time, surveyed both theoretical and experimental particle physicists to track researchers’ attitudes to physics beyond the standard model. The survey results, however, will only get published in January, so I presently can’t tell you more than that. But once the paper is available you’ll read about it on this blog.

The next talk was by Radin Dardashti who warned us ahead that he’d be speaking about work in progress. I very much liked Radin’s talk at last year’s workshop, and this one didn’t disappoint either. In his new work, he is trying to make precise the notion of “theory space” (in the general sense, not restricted to qfts).

I think it’s a brilliant idea because there are many things that we know about theories but that aren’t about any particular theory, ie we know something about theory space, but we never formalize this knowledge. The most obvious example may be that theories in physics tend to be nice and smooth and well-behaved. They can be extrapolated. They have differentiable potentials. They can be expanded. There isn’t a priori any reason why that should be so; it’s just a lesson we have learned through history. I believe that quantifying meta-theoretical knowledge like this could play a useful role in theory development. I also believe Radin has a bright future ahead.

The final session on Tuesday afternoon was the most physicsy one.

My own talk about the role of arguments from naturalness was followed by a rather puzzling contribution by two young philosophers. They claimed that quantum gravity doesn’t have to be UV-complete, which would mean it’s not a consistent theory up to arbitrarily high energies.

It’s right of course that quantum gravity doesn’t have to be UV-complete, but it’s kinda like saying a plane doesn’t have to fly. If you don’t mind driving, then why put wings on it? If you don’t mind UV-incompleteness, then why quantize gravity?

This isn’t to say that there’s no use in thinking about approximations to quantum gravity which aren’t UV-complete and, in particular, trying to find ways to test them. But these are means to an end, and the end is still UV-completion. Now we can discuss whether it’s a good idea to start with the end rather than the means, but that’s a different story and shall be told another time.

I think this talk confused me because the argument wasn’t wrong, but for a practicing researcher in the field the consideration is remarkably irrelevant. Our first concern is to find a promising problem to work on, and that the combination of quantum field theory and general relativity isn’t UV complete is the most promising problem I know of.

The last talk was by Michael Krämer about recent developments in modelling particle dark matter. In astrophysics – like in particle-physics – the trend is to go away from top-down models and work with slimmer “simplified” models. I think it’s a good trend because the top-down constructions didn’t lead us anywhere. But removing the top-down guidance must be accompanied by new criteria, some new principle of non-empirical theory-selection, which I’m still waiting to see. Otherwise we’ll just endlessly produce models of questionable relevance.

I’m not sure whether a few days with a group of philosophers have improved my reasoning – be my judge. But the workshop helped me see the reason I’ve recently drifted towards philosophy: I’m frustrated by the lack of self-reflection among theoretical physicists. In the foundations of physics, everybody’s running at high speed without getting anywhere, and yet they never stop to ask what might possibly be going wrong. Indeed, most of them will insist nothing’s wrong to begin with. The philosophers are offering the conceptual clarity that I find missing in my own field.

I guess I’ll be back.


  1. Excellent review as always Dr. H! Since I am a philosopher I find myself appreciating your appreciation. I am presently re-reading E.J. Lowe's "The Possibility of Metaphysics" (1998) which is specifically directed at metaphysical clarification for the sake of advancing science (in general not physics specifically) another good read for you if ever you had the time. Glad you are enjoying the conference. Hope you have a nice holiday season

  2. Matthew,

    I'm presently reading a book on metaphysics, and while it's not bad I don't find it particularly useful either. So, thanks for the reading recommendation.

  3. time-symmetry in quantum mechanics implies retro-causality Pure reason absent empirical postulates fails. Euclid’s Fifth Postulate is incomplete re maps, Shroud of Turin. Newton bobbled c, h, and k_b. The Equivalence Principle prohibits baryogenesis.

    Breaking time reversal symmetry creates chirality. arXiv:1006.0762, doi:10.1103/PhysRevA.82.043811. Chirality is the strong arrow of time. The Second Law is statistical. doi:10.1038/nature08680; arXiv:hep-ph/0501282, doi:10.1103/PhysRevD.71.057501; arXiv:condmat/0307056, doi:10.1103/PhysRevLett.91.247404; doi:10.1016/0009-2614(90)87240-R

    Matter in excess of antimatter prohibits retro-causality. SUSY and M-theory are not empirical at exact spatial mirror symmetries. Dark matter is Milgrom acceleration, Noetherean leakage of angular momentum via trace spatial chiral anisotropy.

  4. "Should “constern” be a verb? Discuss."

    I don't think so since the word "consternate" already exists for the implied meaning of "constern".

  5. Thanks Sabine! But you did bestow two extra 't's on me!
    Best, Mathias Frisch :)

  6. "Constern" is a great word! Particularly because it sounds so much like "concern" but with more depth of concern-edness. It's like my old boss who used to say he was "flustrated" = "flustered" + "frustrated"

  7. Just as the Standard Model may not be a UV-completion of strong, weak & EM, a theory of quantum gravity might go beyond an effective quantum theory of gravity based on general relativity, but not be UV-complete.

  8. Mathias, Sorry about that, I took back a few 't' :o)

  9. Arun,

    Yes, it might. What I'm saying is that if you want to develop a theory on purely mathematical grounds (well, we all know there's no data, right?) then you need a very convincing mathematical criterion for guidance, and UV-completion is exactly this. The thought being that the lack of UV-completion is how we *know* (can't emphasize this enough) the current theories are incomplete.

    Now if you say I don't care about UV-completion, you can do all kinds of things, fine. But what's your reason to believe it has anything to do with nature? If you do that to construct a testable phenomenological model, fine - these models have a different purpose, they're there to guide theory development all by themselves. But if you allow people to construct theories that neither solve a problem nor are testable, where'd this get us? (That's basically what happened with inflation, and there you can see where it got us.) Best,


  10. That's basically what happened with inflation.....

    Care to elaborate what you mean by that?

  11. B,

    Inflation was invented to solve several finetuning problems. It's debatable whether these are problems that needed solving to begin with - they're clearly not on a mathematical level as rigorous as self-consistency. Either way, inflation doesn't actually solve these problems. But now it's become a business, so people are producing more and more models. There are infinitely many field configurations and potentials that one can think up, and therefore I don't see any end to this model production. Hence I think it's a good example what happens if you take away a quality criterion (in this case, limiting finetuning). It gives people a free pass to do anything, basically (provided you play by the rules of the field). The outcome is a lot of publications with very little lasting scientific value. Best,


  12. Bee, interesting blog. I really enjoy your quirky, slightly 'sarky', and insightful sense of humour - especially in this case as neatly embodied in the fourth paragraph outlining workshop themes. BTW, do you ever respond to Uncle Al's posts?

  13. I am not a science professional, but I have a question that has worried me for a while: Do those of you who rely on mathematics as a proof believe that math is an analogy? That is, that math is an analogous, or symbolic, system that describes reality (like language, maybe)rather than a reality itself?

  14. I am not a physicist or mathematician but I do believe you are correct. Math follows reality, it doesn't determine it.

  15. ...1) Organic chemists respect chiral consequences. Perhaps Bee is amused by my heresies’ trivial testability.
    ...2) Theoretical physics is rigorous, aesthetic, then tuned to observation. String/M-theory quashed falsifiability.
    ...3) Experimental physics as business model observes within theory. 1928 Cox recanted parity violation. 1956 Yang and Lee had Madame Wu.
    ...4) Green’s function removes geometric chirality. Mirror asymmetry has ugly matrices and prolix equations. Empirical error is merely ~10^(-10), prohibiting baryogenesis and requiring dark matter, spraying theory with tiny holes.

    Physics, “Idiot! Now we are both in the hole.”
    Chemistry, “I live in this hole. I know a way out.”


  16. David,

    Thanks. I rarely respond to Uncle Al's posts. He rarely seems to actually want a reply.

  17. Hello Dr. Hossenfelder

    I enjoyed this article. Thank you for your highly informative posts throughout the year. I learn a lot from them.

    Happy holidays.

    Neil Bruce

  18. Bee, Indeed! Since I started reading your blogs a little while back, Uncle Al's posts have usually been statements and not questions! They are very Alice-in-Wonderland-ish, and seem to stem from a supremely hard-core physicist/chemist. They invariably leave me with an underlying sense of questioning that make me think that I really don't know much - or that he is poetically simply 'pulling my leg'! Anyway, I'm looking forward to all your future blogs - they brighten and enlighten my day - I'm glad that I stumbled onto your interesting blog site! Happy Holidays!

  19. Uncle Al's writings always remind me of the postmodern text generator and it's descendants:

    SCIgen - An Automatic CS Paper Generator

  20. An alpha-quartz single crystal is self-similar down to one 0.113 nm³ unit cell, enantiomorphic space groups P3(1)21 right-handed or P3(2)21 left-handed. Every atom is in a 3-fold helix, 3(1) or 3(2). A geometric Eötvös experiment loads four 5-gram P3(1)21 test masses against four 5-gram P3(2)21 test masses, 6.68×10^22 pairs of 0.113 nm³ opposite shoes. All measurable observables exactly cancel, geometry perfectly diverges (anonymous unit-mass point distribution in space. doi:10.1107/S0108767303004161, Section 3ff).

    ...1) The Equivalence Principle fails. Baryogenesis happened. Exact space-time mirror symmetry is flawed. No postulate can resist empirical falsification.
    ...2) The Equivalence Principle holds. Baryogenesis is forbidden. Space-time is exactly mirror-symmetric.
    ...Middle of page, "enantiomorphous pairs"
    ...Exhaustive crystallographic handedness. P3(1,2)21 and P3(1,2) are maxima.

  21. piein,

    I didn't approve several of your comments, not because of the "babe" but because they don't contain any relevant information. I don't care at all what you think I should "admit to myself" or what "But have fun with yourself being wined and dined by the hungry wolves" is supposed to mean, but nobody here will have any benefit from your rambles. Neither, for that matter, do I care whether you think your comments were "reasonable". If I don't think a comment is useful for my readers, I'll not post it. Best,


  22. Dear B, thank you for the considerable time and effort you put into these posts, which are my main line into a critical (but reasonable!) view of, particularly, physics and gravity. They are my favorite read on the internet.

    I can see how a combination of advances using analogue systems, and the work on theory space, could bring fresh ideas and new insights on their validity, but I'm not clear on how these relate to the 'real world'. At some point there has to be predictions that intersect our energy regime, and are accessible to experiment.

  23. Recent particle physics experiments may be telling us there's no SUSY, GUTs, etc.
    (May or may not, for the sake of the argument let's say they do.)

    If all these ideas are wrong may these experimental results be telling us that the initial hypotheses that led to them were wrong too?
    (extension of gauge group, unification/enlargement of gauge groups, hierarchy, naturalness,...)

    If that is the case (and again, it may or may not be), how far back should we look to find a nice solid ground again? Experimentally it is clear which "truths" hold and which do not. But how about the theory? Which theoretical "truths" can we be truly certain about?


  24. wow, took the easy target sentence out of the middle of its context and heaped ridicule on it for lacking content, and finished with a glob of self-glorifying insincerity about the good of your readers. Consider me humiliated. On your side you'd have to literally murder someone or sell crack or something like that to get any lower. I don't personally give a shit, but just to say you will eventually make enemies with one sided humiliations based on a deceit facilitated by a misuse of position. This blog constitutes a tiny little chunk sized baby mushed exposure to power, and yet you fail to resist the temptation to misuse it for personal convenience. That's a catastrophic failure of character. I challenge you to post my whole original statement, and criticize all of it, and allow me to defend. Because from reading this blog I am confident that your knowledge is patchy. Did you switch from mathematics to physics because you were failing or giving up? I think the answer to that is yes.

  25. piein,

    I am posting your comment so that everybody can see the quality of your comments, and will then move on to put you on the blocked list. Good bye,


  26. Vedran,

    Well, yes, we should reconsider these hypotheses. But it isn't happening. Best,


  27. Pfogle,

    Yes, eventually you want to arrive at testable predictions, that's the point. If you use one quantum system to 'compute' what another one does, you can then go and test how well it did. Or, somewhat closer to home for me, you want to test how far an analogy between gravity and (certain types of fluids) holds. Best,


  28. Sabine said, "I’m frustrated by the lack of self-reflection among theoretical physicists"

    It's good to see a researcher say that for the same reason as you've said in other articles; more honest self-reflection would likely hasten the science progress. I believe we may both share a frustration that advancements are unnecessarily slowed by human behaviors that creep into scientific methodology, even though the "behavior creep" is most often unintentional.

  29. Hello Sabine

    Like your comment "This consterns me because, as I explained a decade ago, it’s an unsound abuse of probability calculus. You can’t randomly distribute events that are causally related. It’s mathematical nonsense, end of story"!

    I have the same problem with the use of statistical mechanics in market analysis (human behavior is yoked) but the probability theory makes the assumption the particles are not yoked in anyway; and in biomedical applications where the degrees of freedom (genetic diversity and especially during brain development) are not only unknown but highly variable in the statistical methods used often. I suspect much of the popular statistical outcomes publicized in medicine of A causes B or A is not the cause of B garnered from large scale population studies is flawed. Genomic sciences like the human 1000 genome project reveals huge diversity. Worry, we need more incisive thinking rather than relying on mindless probability calculus machine algorithms in the future. Love to hear your thoughts on this.

  30. rene,

    It's not a topic I know much about, sorry.


COMMENTS ON THIS BLOG ARE PERMANENTLY CLOSED. You can join the discussion on Patreon.

Note: Only a member of this blog may post a comment.