About That Smoking Gun…

Aeon magazine has just published a long piece on the current state of cosmology by Ross Andersen. One focus is on Paul Steinhardt and his claims that the popular multiverse/eternal inflation scenario doesn’t explain what it is supposed to, and is compatible with almost any experimental result. The BICEP2 fiasco, where multiverse proponents first claimed a “smoking gun” vindication from B-modes, then went on to claim that no B-modes was just as good for their theory once they disappeared, is the main topic of the article (for a previous posting about this, see here).

That’s why Steinhardt was surprised to see inflationary theorists clinking glasses when BICEP2 announced a high swirls figure. ‘They declared victory,’ he told me. ‘They said it was smoking-gun proof! Just what they expected!’

But then a few months passed and BICEP2’s interpretation started to look wobbly. In June, Linde told New Scientist that he didn’t like the way BICEP2’s swirls were being treated as a smoking gun for inflation. In July, Guth made similar statements to the Washington Post. Steinhardt was furious. He thought it was flip-flopping. He began to wonder if any data would disturb the serene certainty of inflationary theorists. ‘It was Andre Linde who used the “smoking gun” language in the first place,’ he told me. ‘Now he says it doesn’t make a difference what BICEP2 says. How can it be that not seeing gravitational waves is fine, and then seeing them is a smoking gun, and then not seeing them is fine again?’

Steinhardt explains that the underlying problem is that the underlying problem is an inherently untestable paradigm, compatible with anything:

The theory’s weaknesses can be explained away with the same glib shrug that accompanies the retort: ‘God just made it that way.’

A dominant, infinitely flexible multiverse theory could make it easy not to strain for the next leap forward. It could lead to a chilling effect on new ideas in cosmology, or worse, a creative crisis. Steinhardt thinks we’re already there. ‘Andre Linde has become associated with eternal inflation because he thinks the multiverse is a good idea,’ he told me. ‘But I invented it, too, and I think it’s a horrible idea. It’s an emperor’s new clothes story. Except in that story, it’s a child who points out that the Emperor has no clothes. In this case, it’s the tailors themselves telling us that the theory is not testable. It’s Guth and Linde.’

His hope for how the subject will get saved from itself is with help from philosophers:

‘The outside community isn’t recognising the problem,’ he said. ‘This whole BICEP2 thing has made some people more aware of it. It’s been nice to have that aired out. But most people give us too much respect. They think we know what we’re doing. They take too seriously these voices that say inflation is established theory.’

I asked him who might help. What cavalry was he calling for?

‘I wish the philosophers would get involved,’ he said.

That might help, but I think other ideas are needed…

This entry was posted in Multiverse Mania. Bookmark the permalink.

33 Responses to About That Smoking Gun…

  1. Theo Nieuwenhuizen says:

    I can’t help for longing for the 1930’s, even if it was before my time. People were very cautious with new claims. Pauli felt bad for proposing a new particle – the neutrino.
    Nowadays, if you don’t propose a whole new sector of the standard model (SuSy, mirror matter, …) you won’t be taken serious. Better proclaim smoking-gun inflation or even multiverses (= undecipherable verses).

    What do we really have? Just the standard model particles, with the stable ones being protons, electrons, photons, neutrinos. In the 1930s one would have tried to construct a Universe from that. And one would have hunted the bears-on-the-road like neutrino free-streaming, that seems to be a road block (but is not).

    Great times. I bet we will be forced to return to this level of modesty. And have to wonder how we could get it so wrong.

  2. Jim Given says:

    The philosophers of science have already told us everything we need to know:
    Every domain of Big Science needs a “governing narrative” to provide a basis for discussion and to evaluate research priorities. Such a narrative, a Big Theory, is constructed so as to be non-falsifiable. to provide continuity. So attempts to directly overturn Big Theories, whether Freud’s in psychology, Chomsky’s in linguistics, or inflationary cosmology in this case; are futile. What must happen is for funders and research directors to conclude the dominant Big Theory is no longer a productive source of interesting research directions; and for bright young theorists to work on other paradigms and find success. Perhaps future historians of science will conclude we are already well into this phase. But Big Theories never fail; those with a lifelong vested interest in them retire or die and are replaced by theorists with a different perspective. (This is all vintage Imre Lakatos.)
    *** May you live in interesting times. ***

  3. Jim Given: I think it’s a little more complex, and a great deal more interesting, than that. Think about Einstein and entanglement. He hated its implication of irreducible indeterminacy because it conflicted with what you are calling a “governing narrative,” which for Einstein was the idea that natural law is deterministic and local. To fight this spooky implication, Einstein, Podolsky, and Rosen wrote a paper which has become extremely influential and productive, even though it is wrong.

    I guess there’s a saying of some sort lurking in there: “wrong enough to be useful” or something like that.

    Anyway, scientists of different views are, in my judgment, at least the good ones, involved in something more like a dialectic than a die-off.

    Also, I think the final arbiters of how this all works are the scientists themselves, not philosophers. The scientists have implicit philosophies, which I call “the philosophy of scientists” as opposed to “the philosophy of science,” which are usually pretty vague and not formulated to the standards of philosophical papers, but on the other hand, the empirical successes of science warrant that there is something important going on in “the philosophy of scientists.”

  4. Steinhardt is disingeniously conflating eternal inflation and inflation. Inflation is not necessarily eternal (many models, including ones consistent with the data, such as Higgs inflation, do not lead to eternal inflation), and does not in any way imply the multiverse. It’s true that some inflationary theorists are promoting the chain of inference inflation -> eternal inflation -> multiverse. But the criticism should be directed towards the theorists, not inflationary theory. Instead, Steinhardt is adopting this false chain of reasoning to promote his own competing ideas.

    PS. The first person to suggest inflation as a solution to the horizon problem was not Guth, but Kazanas: http://adsabs.harvard.edu/doi/10.1086/183361

  5. Thomas Larsson says:

    Theo, I think the situation today may be quite similar to the 1930s. At that time people had discovered that QM was incompatible with electromagnetics, and many people apparantly thought that the only way to solve that problem was to postulate new physics – I have seen claims that Oppenheimer thought so. Alas, we all know that the solution was not new physics, but a new way to look at old physics – renormalization. Perhaps there is a lesson there for our time, when everything but SM + GR looks increasingly unlikely.

  6. JG says:

    The role of philosophers belongs to pre-1950s and even much earlier when not many scientists took cosmological models serious.

    Since then we have a remarkably good scientific cosmological model – people should (re-)read Weinberg’s The First Three Minutes from 1977(!) to remind just how good a scientific theory we have even without Inflation, Dark Matter/Energy and other more recent theoretical suggestions. (Weinberg updated the book in 1993, before the accelerating expansion had been discovered but he discussed many of the other newer ideas)

    None of these additions since the late 1970s are of a philosophical nature, they are deep mathematically constructed ideas that require a high degree of understanding of a lot of known physics.

    Philosophers can only muddy the waters, we need brilliant mathematical/physical insights building on a solid understanding of the existing theory. We may even already have discovered the true model.

  7. Art Brown says:

    It’s a pity the article didn’t include any responses from Linde and/or Guth, given Steinhardt’s statements about them.

  8. Dave Miller in Sacramento says:

    Perhaps someone more knowledgeable than I can let me know if my impression of inflation is close to the mark:

    My impression is that inflation is an interesting idea, which might explain some interesting features of the universe (large-scale homogeneity, near flatness, etc.), which is consistent with existing cosmological data, and which perhaps has a little bit of observational data that supports it but that is inconclusive.

    Is that pretty much it?

    Incidentally, I was at SLAC when Guth did his initial inflation work there, and I went to hear one of Alan’s early talks on the subject: it sounded interesting and still sounds interesting (the Big Bang as the ultimate beginning was always a bit dodgy), but I am curious as to whether it can ever be confirmed or disproven.

    Dave Miller in Sacramento.

  9. Dave Miller in Sacramento says:

    Jim Given wrote:

    The philosophers of science have already told us everything we need to know: Every domain of Big Science needs a “governing narrative” to provide a basis for discussion and to evaluate research priorities. Such a narrative, a Big Theory, is constructed so as to be non-falsifiable. to provide continuity.

    Except… a lot of those “governing narratives” have been falsified: the geocentric theory, the immutability of species, the absoluteness of space and time, the immovability of the continents, etc.

    So… the philosophical theory you mention has in fact itself been falsified!


  10. Neil says:

    Send in the philosophers? I doubt that would help from what I have read so far (see Richard Dawid’s book on “post-empiricism”.) The multiverse, like free will, offers philosophers a dream subject–something that can be debated endlessly without fear of ever reaching resolution.

  11. Aleksandar Mikovic says:

    The obvious solution to the multiverse problem is to find a quantum gravity theory different from string theory, such that it allows inflation. One example is loop quantum cosmology.

  12. Dave Miller in Sacramento:

    The major success of inflation is not explaining the spatial homogeneity, isotropy and flatness of the background (although the last was not obvious at the time inflation was proposed), but accounting in detail for the perturbations in terms of quantum fluctuations, first discussed by Mukhanov and Chibisov.

    The perturbation calculation is non-trivial, and it correctly predicted that the primordial perturbations are a) close to scale-invariant, b) dominantly scalar, c) adiabatic, d) statistically homogeneous and isotropic and e) Gaussian. These are real predictions, made before the observations were in. No other proposal has been able to account for them as well as inflation, even after the fact.

    So there is a lot of support for inflation. Whether one regards inflation as proven is a matter of taste. There are hundreds of inflationary models (many of which differ from one or more of predictions given above), and I would prefer to have more understanding of the physics, or observation of the wavelength dependence of the gravitational waves (see below), before saying that we know that inflation occurred.

    As for the smoking gun, the BICEP2 results were misrepresented from the beginning as “first direct evidence of cosmic inflation”, because there was a lot of evidence already, and gravitational waves are not more direct than scalar perturbations. The simplest models of inflation predict a relationship between the amplitude of the scalar perturbations, amplitude of the gravitational waves and the wavelength-dependence of the gravitational wave amplitude. I think that could be called a ‘smoking gun’, in the sense that if it was measured, I think we could say that inflation happened, even without further understanding of the models.

    Please note that even if something is a smoking gun (definite evidence for A), not finding it does not imply that A is not correct.

  13. Alex says:

    @Syksy Räsänen

    “The simplest models of inflation predict a relationship between the amplitude of the scalar perturbations, amplitude of the gravitational waves and the wavelength-dependence of the gravitational wave amplitude. I think that could be called a ‘smoking gun’, in the sense that if it was measured, I think we could say that inflation happened, even without further understanding of the models.”

    But a parameter was measured, and it was found to conflict with the simplest models of inflation. Yet, all this measurement has done was to constrain which models of inflation could possibly work. Therefore I really don’t think this could reasonably be called a smoking gun. Again, all it did was to rule out models of inflation that could not possibly be true.

  14. Alex says:

    @Syksy Räsänen

    I’ve put a very rough argument together to put this idea that this observation would be a “smoking gun” on a slightly more objective footing.

    I think its reasonable to assume that an observation being a “smoking gun” means that this observation must carry some kind of information about the theory we are trying to evaluate. Namely whether it be true or false. The key parameter we’re measuring here would be the scalar-to-tensor ratio denoted by r. The measurement of r should tell us whether our theory is true or false. Denote these outcomes by \omega_r. \omega_r = 0 or 1 for false and true. They key question here is what is the probability that our theory (inflation) is false given some observation of r. I think its quite reasonable to assume (given the above discussion) that P(\omega_r = 0) ~ \epsilon, where \epsilon is a very small number, approaching zero. This quantifying the fact that no matter what value of r we observe, inflation is correct. Hence, P(\omega_r = 1) ~ 1 – \epsilon.

    What is the information given by this measurement which you characterize as being of the “smoking gun” nature. The information entropy is E(I(\omega_r)) = -\sum_{\omega_r}(P(\omega_r)log(P(\omega_r))) = -\epsilon*log(\epsilon) – (1-\epsilon)*log(1-\epsilon) ~ 0.

    Hence, no real information is conveyed given you characterization of this experiment. Hence we do not have a smoking gun.

  15. Alex:

    1. BICEP2 only claimed to have measured the amplitude of gravitational waves, not the dependence of the amplitude on wavenumber. So even if correct, their results would not have been what I called a smoking gun above.

    2. The BICEP2 results, if correct, would not have been in conflict with the simplest models of inflation, though there was some tension. See http://arxiv.org/abs/1405.1390 (though their accounting of the tension is a bit simplistic).

    3. Let us not confuse the concepts “smoking gun” and “test”. A is a test for X, if X implies A. Proving A does not prove X, but disproving A disproves X.

    A is a smoking gun for X, if A implies X. Proving A proves X, but disproving A does not disprove X.

    Seeing gravitational waves that satisfy the consistency condition I referred to above can, in my view, be regarded as a smoking gun for inflation, because it relates three independent, non-trivial, quantities in a way that no other theory has (as far as I know) has been able to reproduce.

    However, it cannot be regarded as a test, because even if the relation holds, the gravitational wave amplitude depends on the model (and can be below the sensitivity of near-future experiments), and there are many models of inflation in which the relation does not hold.

  16. Alex says:

    Syksy Räsänen:

    I’m sorry I think I read part of your comment too quickly. Yes, I agree with point 1) above.

    But as for this part:

    “However, it cannot be regarded as a test, […] and there are many models of inflation in which the relation does not hold.”

    Am I reading this right? You’re saying that one of the reasons it can’t be regarded as a test is *because* there are many models of inflation in which this relationship between wave number and amplitude doesn’t hold anyway?

  17. RM says:

    I think philosophers would be of help to Steinhardt – one of the basic lessons one learns in philosophy is that if A implies B, then Not A does not imply Not B.

  18. Alex says:

    Syksy Räsänen:

    To sum up:

    The alternative test you are proposing seems to be on the same footing as the amplitude measured by BICEP2 in terms of its ability to determine the validity of inflationary theory. If there is a specific relationship between wave number and amplitude, inflation is right. If there isn’t this specific relationship, inflation is right. The problem isn’t with your test, its with inflationary theory itself, its not uniquely defined.

    Also, I disagree with your description of the “smoking gun” term. If an experiment is going to provide me with conclusive evidence that a theory is correct, there ought to be an experimental outcome that would prove the theory incorrect. In other words, the experimental outcome should give actual information (see my above post).

    I think we’re probably going to have to agree to disagree.


  19. Unlike Steinhardt claims, there’s nothing extraordinary about the situation with regard to gravitational waves. This distinction of test vs. smoking gun is common to all cases where there are several models whose predictions overlap in the area where predictions have been made, but which have different predictions outside.

    For example, consider direct detection experiments of dark matter, sensitive to a certain range of mass and interaction cross section. If events are seen that are consistent with dark matter (so that they are consistent with other observations, and backgrounds, systematics etc. are ruled out), this can be regarded as a smoking gun. If they are not seen, this does not rule out dark matter, because its mass and cross section may be out of the range probed by the experiment. This does not mean that the experiment does not give us information about dark matter (it narrows the possibilities).

  20. Andre says:

    Steinhardt criticizes inflation, because if this theory is confirmed, then his own work on the cyclic universe will remain useless. All simple.

  21. Peter Woit says:

    Surely his own work is a motivation, but I think it’s not unreasonable for him to argue that it’s not a fair scientific competition between ideas when one side gets to claim victory no matter what the experiments say. If you take a look at the reaction after the BICEP2 results were announced, see
    you had
    1. “Paul Steinhardt, a theoretical physicist at Princeton University in New Jersey and an originator of the cyclic theory, agrees that — if the BICEP2 findings are confirmed — his theory is now dead”
    although he did say he would look for a cyclic model that would agree with BICEP2.
    2. Linde on the other hand, reacting to incompatibility of BICEP2 with popular string inflation schemes: “There is no need to discard string theory, it is just a normal process of learning which versions of the theory are better,”

  22. Alex says:

    Syksy Räsänen:

    IF you are already convinced inflation ( or dark matter) indeed happened then yes these experiments provide information in the form of narrowing down which specific parameter values in the theory are true.

    But I maintain that these tests provide no information as to whether the theory of inflation itself (for any parameter value) is true. There will always be a parameter value to fit ANY observation. So we can’t “say that inflation happened” based on the results of BICEP2 or the experiment you proposed. But if we assume inflation happened then we can say which version of it happened based on the results of these experiments.

    Ok, the prosecution rests.

  23. Dave Miller in Sacramento says:

    Alex and Syksy,

    Thanks for your informative back-and-forth on inflation.

    Sysksy, you wrote:

    The major success of inflation is… accounting in detail for the perturbations in terms of quantum fluctuations, first discussed by Mukhanov and Chibisov.

    …it correctly predicted that the primordial perturbations are a) close to scale-invariant, b) dominantly scalar, c) adiabatic, d) statistically homogeneous and isotropic and e) Gaussian….

    … There are hundreds of inflationary models (many of which differ from one or more of predictions given above)

    That was sort of my impression, and it strikes me as suggestive, but very, very far from conclusive. After all,, isotropy and Gaussian distributions are, for example, rather common features of phenomena in the real world!

    Having known Alan when he started working on inflation, I have a vested interest in hoping that it is eventually proven to be true: kinda cool to be able to say “I was there when…” (Similarly, I knew Schwarz, Polchinski, and Susskind when I was a student, so it would be kinda cool for me if string theory works out.)

    But… when there are, as you say, “hundreds of inflationary models” that give different results… well, if the theories of Planck, Rutherford, Bohr, et al. had been that rubbery, I think they would have been a good deal less convincing.

    Thanks again for the back-and-forth.


  24. srp says:

    Clearly a smoking gun and a crucial test are not the same things, as Syksy points out. A smoking gun in the suspect’s hand would be powerful evidence of the suspect’s guilt, but the absence of the smoking gun would not be enough to exonerate him. It is perfectly logical to posit such “one-way” tests that are not crucial but could still clinch the truth of a theory if they were passed. (In the same way, we might rule out a suspect if a test turned out one way but not convict him if it turned out the other way, e.g. if we see him far away from the victim on a video time-stamped at the time the crime occurred he’s in the clear, but failing to find such a video wouldn’t make him guilty.)

    On the other hand, if we theorized that the murderer must come from a vaguely defined set of over one million people and claimed that our theory was useful only because a smoking gun might turn up in one of their hands, we might not be hired back as consulting detectives.

  25. Pingback: Paul Steinhardt’s remorse, Popperianism and Beauty-Contest | The Great Vindications

  26. Art Brown says:

    Syksy Räsänen,
    Thanks for the reference to the illuminating Audren et al. analysis of the tension between BICEP2’s original paper and Planck. I realize this topic is now Overtaken By Events, but:
    1) Can you elaborate a bit on your comment that “their accounting of the tension was a bit simplistic”?
    2) What am I to make of the conclusion in BICEP2’s original release (1403.3985v1, 2, & 3): “These high values of r are in apparent tension with previous indirect limits based on temperature limits …”?
    Thanks again.

  27. Calvin Marshall says:

    Dr. Woit,

    As I understand it from Vilenkin’s book ‘Many Worlds in One’, the mechanism for eternal inflation is a scalar field: areas of false vacuum decaying into areas of true vacuum. But as Alexander Vilenkin notes in ‘Many Worlds in One’, there is no direct evidence for the existence of scalar fields in nature: “Another important question is whether or not such scalar fields really exist in nature. Unfortunately, we don’t know. There is no direct evidence for their existence.” (p. 61). Reading this as a layman, this is what I take away: no scalar fields, no eternal inflation. There’s no direct evidence for scalar fields, therefore there’s no direct evidence eternal inflation. Is this a correct understanding?

    Thank you,
    Calvin Marshall

  28. Peter Woit says:

    Calvin Marshall,
    We do now have evidence for a scalar field, the Higgs field. It’s very unclear though whether the Higgs field can be used to give you inflation. The models cosmologists mostly work with use a postulated scalar field to drive inflation, one which has no other observable effects.

  29. Art Brown:

    1) In the paper, the authors find the confidence levels for Planck and BICEP2 separately and look at their overlap. They say that since the overlap is in the region between one and two sigma, there is no significant tension.

    It would be more appropriate to do one or more of the following: 1) ask the Bayesian question of what is the likelihood of the BICEP2 observations given the Planck observations (assuming the simplest inflation model), 2) consider the goodness-of-fit of the model to the full dataset, or 3) see how much the fit is improved by the addition of new parameters.

    All this, of course, assuming the (wrong) interpretation of the B-modes as being due to gravitational waves.

    2) The part you quote is correct: the results were “in apparent tension” (which is different from conflict). For caveats in that BICEP2 statement of tension, see the abstract of the Audren et al paper.

  30. Art Brown:

    I realised now that your question is probably related to the title of the Audren et al paper, “BICEP2 and Planck are not in tension”. I would say that their analysis shows that the tension is not as large as some thought (with the caveats above), not as saying that there is no tension, but this is a matter of taste.

  31. Art Brown says:

    Syksy Räsänen, Thank you very much for the additional explanation. On a separate note, I see you have updated your cosmology notes and look forward to having a go at them.

  32. Art Brown:

    Let me know of any typos!

  33. Hank Bolden says:

    In the New York Times dated June 5, 2015:

    “Crisis at the Edge of Physics” DO physicists need empirical evidence to confirm their theories?


Comments are closed.