Fake News of the Holographic Universe

It has been a very busy day today but I thought I’d grab a few minutes to rant about something inspired by a cosmological topic but that I’m afraid is symptomatic of malaise that extends far wider than fundamental science.

The other day I found a news item with the title Study reveals substantial evidence of holographic universe. You can find a fairly detailed discussion of the holographic principle here, but the name is fairly self-explanatory: the familiar hologram is a two-dimensional object that contains enough information to reconstruct a three-dimensional object. The holographic principle extends this to the idea that information pertaining to a higher-dimensional space may reside on a lower-dimensional boundary of that space. It’s an idea which has gained some traction in the context of the black hole information paradox, for example.

There are people far more knowledgeable about the holographic principle than me, but naturally what grabbed my attention was the title of the news item: Study reveals substantial evidence of holographic universe. That got me really excited, as I wasn’t previously aware that there was any observed property of the Universe that showed any unambiguous evidence for the holographic interpretation or indeed that models based on this model could describe the available data better than the standard ΛCDM cosmological model. Naturally I went to the original paper on the arXiv by Niayesh Ashfordi et al. to which the news item relates. Here is the abstract:

We test a class of holographic models for the very early universe against cosmological observations and find that they are competitive to the standard ΛCDM model of cosmology. These models are based on three dimensional perturbative super-renormalizable Quantum Field Theory (QFT), and while they predict a different power spectrum from the standard power-law used in ΛCDM, they still provide an excellent fit to data (within their regime of validity). By comparing the Bayesian evidence for the models, we find that ΛCDM does a better job globally, while the holographic models provide a (marginally) better fit to data without very low multipoles (i.e. l≲30), where the dual QFT becomes non-perturbative. Observations can be used to exclude some QFT models, while we also find models satisfying all phenomenological constraints: the data rules out the dual theory being Yang-Mills theory coupled to fermions only, but allows for Yang-Mills theory coupled to non-minimal scalars with quartic interactions. Lattice simulations of 3d QFT’s can provide non-perturbative predictions for large-angle statistics of the cosmic microwave background, and potentially explain its apparent anomalies.

The third sentence (highlighted) states explicitly that according to the Bayesian evidence (see here for a review of this) the holographic models do not fit the data even as well as the standard model (unless some of the CMB measurements are excluded, and then they’re only slightly better)

I think the holographic principle is a very interesting idea and it may indeed at some point prove to provide a deeper understanding of our universe than our current models. Nevertheless it seems clear to me that the title of this news article is extremely misleading. Current observations do not really provide any evidence in favour of the holographic models, and certainly not “substantial evidence”.

The wider point should be obvious. We scientists rightly bemoan the era of “fake news”. We like to think that we occupy the high ground, by rigorously weighing up the evidence, drawing conclusions as objectively as possible, and reporting our findings with a balanced view of the uncertainties and caveats. That’s what we should be doing. Unless we do that we’re not communicating science but engaged in propaganda, and that’s a very dangerous game to play as it endangers the already fragile trust the public place in science.

The authors of the paper are not entirely to blame as they did not write the piece that kicked off this rant, which seems to have been produced by the press office at the University of Southampton, but they should not have consented to it being released with such a misleading title.

Advertisements

28 Responses to “Fake News of the Holographic Universe”

  1. Bryn Jones Says:

    I’d be cautious about blaming the authors for some of the hype associated with a research article.

    Years ago some research I was involved with was published in Nature. I started getting telephone calls and e-mails from news outlets. I had done nothing to promote the work, apart from consenting to a reasonable press release from the university press office when the office approached me for one. Reports started appearing all over the place. Some people somewhere were hyping the work. It wasn’t me and probably wasn’t my collaborators.

    So the people hyping research may well be university press offices and journals without the full knowledge or consent of the researchers.

    • The question is whether the terms and conditions of employment give researchers the right to clear all publicity.

      • Bryn Jones Says:

        Yes, and I doubt they do.

        Another question, which may be more relevant still, is whether researchers know what university press offices are saying about their work. Researchers may approve most formal press releases, but they may know nothing about what is said by press offices in telephone calls and e-mail messages to the media. The worst hyping may even be done by press offices of collaborators’ universities.

    • telescoper Says:

      I would be inclined to agree, and said as much at the end of the post. However, in this case we do have a comment from one of the authors of the paper defending the press release. You can make your own mind up what you think of the response.

  2. nafshordi Says:

    Sorry @telescoper, but it seems that you are the one propagating “fake news” here:

    1-Everybody agrees there is substantial evidence for LCDM model.
    2-We show that evidence for LCDM and holographic cosmology are comparable, where they both can be trusted.

    1+2 then implies that “there is substantial evidence” for “holographic cosmology”.

    • telescoper Says:

      I disagree. What you’ve shown is that the evidence doesn’t favour your theory over LCDM.

      • nafshordi Says:

        So what? The point is to explain the data, not to beat the Taylor expansion of power spectrum

    • telescoper Says:

      Two suspects are tested for DNA match to a sample found at the scene. Both match, but Suspect A matches slightly better than Suspect B.

      Would you say that there is “substantial evidence” that B is guilty?

    • Ch4rli3_G0rd0n Says:

      hi, pardon me for hijacking the thread, i’m curious what the implications of your research are relative to Verlinde’s emergent gravity theory and its modified/limited holographic theory, if there are any. Thanks!

  3. Without having gone into the details of the paper enough to understand the calculations, and only looking at the abstract and bit of the paper showing the relevant portions, I disagree with telescoper.

    My understanding from the quick look is that the perturbative framework they have used for calculation breaks down at low multipoles (something they estimate loosely to be at ell ~ 30). Hence they compare l > 30 to observations, since the theoretical calculation is not valid, and not because the data l < 30 would make the comparison look bad. This is perfectly sensible (assuming that they did not just make up the estimate because the alternative would look bad).

    If I only had the tools of linear perturbation theory in standard cosmology, would it not be ok to disregard high k matter power spectrum observations when confronting the theoretical computation with power spectrum estimates from galaxy surveys and comparing with some SCDM like model? This seems to be exactly analogous to what they are doing.

    • telescoper Says:

      Yes, that’s right, except that they are discarding long wavelength modes rather than short.

      The bigger problem is that they’re not really testing the right things. One could easily make up an inflationary model that produces the power spectrum in their equation (2) by simply choosing an appropriate potential.

      To provide really substantial evidence for the Holographic Universe they need to show it does something that standard cosmology can’t.

  4. Kostas Skenderis Says:

    This is not another inflationary model. The entire framework is different. I don’t think there is a choice of potential that can reproduce the predictions of this model (power spectra and non-gausianities). This model is based on perturbative QFT in three dimensions and its validity is controlled by the size of the coupling constant. As it turns out, the coupling constant becomes large at long wavelengths (roughly l<30), so perturbation theory cannot be used to make predictions there. In this regime, one needs to compute the observables non-perturbatively. When these results become available (and this is in progress) then one can meaningfully compare the predictions of this holographic model against LambdaCDM also at l30 modes and in that range both models do equally well. Neither the paper nor the press release claimed that the holographic model is preferred over LambdaCDM. It actually explicitly says that there is equal evidence for both models. Where is the controversy?

    From the theoretical perspective the holographic framework has a lot more to offer than standard cosmology. In standard cosmology there are UV and fine-tuning issues. In contrast the holographic model is free of such problems.

    • telescoper Says:

      I know it’s not an inflationary model, I’m just saying that I don’t think you can argue that the power spectrum is probative.

      Since there’s a simple one-to-one correspondence between the potential in inflationary models and the perturbation spectrum it seems there must be a potential that can reproduce any spectrum. It may not be a slow-roll model, of course, but that’s not necessarily a problem.

      If you can make valid predictions about other properties then this model may prove interesting, but claiming that there’s strong evidence for it from the power spectrum alone is unjustified, in my opinion.

      • We have already worked out all predictions that could potentially be compared against CMB observations (in the near or not so near future): the tensor power spectrum and all bispectra. See arXiv:1010.0244 for an early account of the observational signatures, though a lot more work was done on non-gaussianities since that review was written. For example, the scalar bispectrum is of the exact factorizable equilateral form with f_{NL}=5/36.

        If you work out a potential that would reproduce the holographic scalar power spectrum then the gravitational theory with this potential will have very different tensors than the holographic model. This is a truly non-geometric model.

        Clearly, the power spectrum in general is not sufficient by itself to discriminate between theories. Based on the available data however Bayesian evidence shows that this model, within its regime of validity, is equally likely as LambdaCDM. For me this is substantial evidence that this model should be taken seriously, as seriously as LambdaCDM.

      • telescoper Says:

        What’s the tensor:scalar ratio in your model, or can’t you predict that at low l?

      • telescoper Says:

        Can I ask another question please?

        In the paper you compare your model with a LCDM that has a free parameter for a running spectral index, rather than the standard minimal model.

        If I recall correctly the CMB data do not need the running spectral index, i.e. I think that a model with that extra parameter is disfavoured with respect to the minimal model by the usual Bayesian model selection (Ockham’s razor).

        Did you try doing a model comparison of your model with the minimal LCDM model?

      • Regarding the questions in the thread below:

        The formula for r, Equation (10), was derived using perturbation theory and as such it is not valid at low l. One needs a new computation to obtain what r is at low l (and such computation is in the pipeline).

        In the paper we compared the holographic model both with the minimal LambdaCDM and also with LambdaCDM with running.
        As one can see from Fig 3 LambdaCDM with running is disfavoured relative to both minimal LambdaCDM and the holographic model, while the difference in Bayesian evidence between minimal LambdaCDM and holographic model is insignificant, with one of the other model being marginally preferred depending on the priors.

      • telescoper Says:

        Thanks for that. I didn’t have access to the paper when I made that comment. I had mis-remembered that your comparison was with the non-minimal LCDM model.

  5. For the moderator:

    The sentence:

    When these results become available (and this is in progress) then one can meaningfully compare the predictions of this holographic model against LambdaCDM also at l30 modes and in that range both models do equally well.

    should be

    When these results become available (and this is in progress) then one can meaningfully compare the predictions of this holographic model against LambdaCDM also at l30 and in that range both models do equally well.

    I would appreciate if this is fixed before my previous message is posted.

    • telescoper Says:

      I’m confused by both versions of this.

      • Here it is again:

        The comparison of the two models is only meaningful for multipoles greater than 30, because only in this range the perturbative computation that led to the holographic power spectrum is valid. In this range both models fit the data equally well.

        For multipoles less than 30, one needs a new (non-perturbative) computation of the holographic power spectrum and when this result becomes available it would make then sense to compare this holographic model with LambdaCDM over the entire range of multipoles.

      • telescoper Says:

        Right, that makes sense.

  6. For the moderator:

    It seems the text editor changes the sentences. The two versions became identical after I posted them.

  7. > except that they are discarding long wavelength modes rather than short.

    I think they discard long wavelengths because these dual theories are perturbative in the short wavelength limit.

    > The bigger problem is that they’re not really testing the right things. One could easily make up an inflationary model that produces the power spectrum in their equation (2) by simply choosing an appropriate potential.

    True, in all two point statistics, the direct quantity of interest is always going to be the power spectrum, and not the mechanism created it. But while a potential can probably be written down for any power spectrum, that might be a little less interesting than starting with some basic principle.

    Then again, it would seem that they are exploring a representative of a new class of models rather than a model which we believe for other reasons. In that respect, it seems that this is more an order of magnitude calculation, and it would seem that the use of Bayesian evidence (while perfectly correct) which is usually an expensive calculation is not worth the effort.

    > To provide really substantial evidence for the Holographic Universe they need to show it does something that standard cosmology can’t

    To be clear, I completely agree that there is no evidence we should drop the standard cosmological model in favour of this model (and the paper did not claim anything of the kind). I think the point of the paper is that the power spectrum in such a model has a Bayesian evidence which is similar to standard cosmology in the calculated range and is not bluntly ruled out, and that is quite interesting, but not exactly in a world changing kind of way. My disagreement with your statement had more to do with whether it was kosher to drop the low multipoles.

    Now coming back to the news piece which triggered your post, I think this may be called ‘substantial’ Bayesian evidence, but not substantial evidence for holographic cosmology over standard cosmology. And I would agree with you that the headline sounds more like the latter rather than the former. In that sense, I would agree that I was not very clear about my disagreement.

  8. […] few days ago I wrote a very sceptical post about an alternative to the present standard cosmological which is called the holographic universe. […]

  9. Davide Castelvecchi Says:

    I would like to point out that there is a difference between “news item” and “press release”: Eurekalert is a portal of press releases, and this one came straight from a university, not from a news source.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: