Polchinski on the Landscape

At the recent Solvay conference there was extensive discussion of the Landscape, and I’ve already discussed here Michael Douglas’s write-up of his talk on the subject, entitled Understanding the Landscape. Now Joe Polchinksi’s rapporteur talk on the subject has appeared; it’s entitled The Cosmological Constant and the String Landscape. Polchinski is essentially making the same argument as Susskind makes in his recent book, and his argument has pretty much the same problems that Susskind’s has, which I discussed in detail in an earlier posting. However, his article is written for physicists, not for the general public, so he is making a much more technical version of the argument. It’s probably the best version of the case for the string theory anthropic landscape available, so worth reading carefully.

Polchinski begins by explaining why the cosmological constant problem is difficult. He divides possible solutions to it into ones where the CC is fixed, and ones where it is adjustable. The problem with the idea that the CC is fixed and calculable is essentially that all the contributions to the vacuum energy that we know about give values of the CC that are far too large. These include things like fermion loops and the Higgs potential in the standard model, supersymmetry breaking in supersymmetric extensions of the SM, and Planck scale effects in theories of quantum gravity. He goes on to explain why it is difficult to try and modify the theory of gravity so that it won’t couple to these sources of vacuum energy.

There is a wide range of ideas about how to select a small CC in theories where the CC is adjustable, and Polchinski describes several of them and what problems they have. Some of these ideas naturally explain a small CC in an empty universe, but not in our matter-filled universe. Finally he ends up with the anthropic explanation: the CC is small because if it wasn’t we wouldn’t be here. He does note that:

Of course, the anthropic principle is in some sense a tautology: we must live where we can live.

this carries a footnote comparing the anthropic principle to Darwin’s theory of natural selection:

Natural selection is a tautology in much the same sense: survivors survive. But in combination with a mechanism of populating a spectrum of universes or genotypes, these ‘tautologies’ acquire great power.

Susskind makes the same sort of claim that the anthropic landscape is much like the theory of evolution, and I think this is extremely dangerous and unwise. There is a mountain of scientific evidence for the theory of evolution and none for the anthropic landscape. It is very important that this distinction be made, and trying to blur the difference between these two very different situations is not something a scientist should be doing at a time when science in general and the theory of evolution in particular is under attack by the forces of the religious right. When I was in Niger one of my travel-mates was a creationist who was generally annoyed at what he saw as the arrogant way scientists dismiss his view of the world. We discussed cosmology, and I tried to tell him a bit about inflation and what aspects of the universe it was supposed to explain. In response he asked me why he shouldn’t just go with his preferred explanation: it was all the doing of the “Big Kahuna”, as he called the deity. I tried to explain about the scientific method: your theory is supposed to make distinctive predictions that you can go out and check by making observations. I think I had some success in getting this idea across, but I don’t see any way I could have defended to him something like the anthropic landscape. It’s not legitimate science since it makes no real predictions that you can use to see if the idea is right or not. What’s at issue here is the credibility of science itself, and physicists who care about this credibility should not be claiming that an idea like the anthropic landscape has the same status as heavily tested and verified theories like that of evolution.

Polchinski goes on to repeat the analogy with evolution a bit later, and I actually don’t understand at all what point he is making here, when he argues that maybe only the anthropic principle determines the CC:

Thus we should seriously consider the possibility that there is no other selection mechanism significantly constraining the cosmological constant. Equally we should not stop searching for such a further principle, but I think one must admit that the strongest reason for expecting to find it is not a scientific argument but a psychological one (footnote): we wish fundamental theory to be as we have long assumed it would be.

and the footnote is:

Again, the Darwinian analogy is notable.

Personally I’m agnostic about whether the CC is computable from first principles or not. As a scientist, one’s job is to come up with theories and extract predictions from them to see if they are right. If you have a theory that says the CC is not computable, that’s fine, but then you have to just forget about the CC and find something that your theory does predict so you can test it. You can’t go around claiming that the fact that your theory is compatible with any value of the CC is somehow some sort of scientific success and evidence for the theory. The problem with a theory where all values of the CC are a priori equally likely is that it’s vacuous (as far as the CC is concerned). Its implications are exactly the same as throwing up your hands and saying “I have absolutely no idea what determines the CC, it could be anything”.

The fact that we are here and observe galaxies does put constraints on the CC, and Polchinski would like to make much of this. He’d like to claim that the anthropic landscape predicts that the CC is a random variable, and then, given the fact we see galaxies, it would have an expectation value about an order of magnitude higher than its observed value. He defends the theory against this mismatch by quoting Galileo’s defense that his theory might be inaccurate but was a lot better than Aristotle’s, then writes:

This order of magnitude may simply be a 1.5 sigma fluctuation, or it may reflect our current ignorance of the measure of the space of vacua

One problem with all this that Polchinski doesn’t mention at this point is that the anthropic constraint is not on the CC but on a combination of the CC and Q, the normalization of the primordial temperature fluctuations. Assuming both are random variables, the observed CC is way off what one expects. One can deal with this by just assuming that the CC is a random variable, but Q isn’t for some reason. This gets into the fundamental problem with the string theory anthropic landscape: it doesn’t just not predict the CC, it doesn’t predict anything at all. As far as one can tell, it’s consistent with just about anything. It doesn’t make any predictions, so it’s really not a legitimate scientific theory at all. One can try and claim that it really is a scientific theory, and that it does predict something: we are at some randomly chosen (according to a not yet understood measure) point in the landscape compatible with our existence. The problem with this (as Polchinski notes) is that there is a long list of properties of the world that appear to be rather special and statistically highly disfavored by any likely measure: the theta-angle is very small, proton lifetime is very long, number of generations is small, etc., etc. If one actually took the string theory anthropic landscape seriously as a theory, one would have to abandon it in face of these falsifying observations.

The crucial question for whether the anthropic landscape is science or not is whether it makes any testable predictions. Polchinski doesn’t at all address the question of whether further study of the landscape will lead to any prediction of anything, presumably because like everyone else he has no plausible idea for how this could come about. He does claim that Weinberg’s 1987 arguement makes 5 successful predictions or postdictions:

The anthropic argument is not without predictive power. We can identify a list of post- or pre-dictions, circa 1987:

1. The cosmological constant is not large.

2. The cosmological constant is not zero.

3. The cosmological constant is similar in order of magnitude to the matter density.

4. As the theory of quantum gravity is better understood, it will provide a microphysics in which the cosmological constant is not fixed but environmental; if this takes discrete values these must be extremely dense in Planck units.

5. Other constants of nature may show evidence of anthropic constraints.

Calling these successful predictions seems to me a huge stretch. 1. follows from a tautology, 2. and 3. are the same “predictions” I get by saying I have no idea what is going on here (including having no good reason to believe the CC is zero). 4. isn’t an experimental prediction at all, and 5. is so vague as to be completely meaningless.

Polchinski ends up his defense of the landscape by quoting Dirac:

One must be prepared to follow up the consequences of theory, and feel that one just has to accept the consequences no matter where they lead.

The problem with this is that Dirac undoubtedly didn’t have in mind the idea that if your theory has no experimental consequences, you should accept the idea that you can’t ever predict anything. Obviously, you should give up on your theory at that point, and this is what Polchinski and others show no signs of being willing to even consider. Back in 1998, in lectures at the SLAC Summer Institute, he wrote:

On Lance Dixon’s tentative outline for my lectures, one of the items was ‘Alternatives to String Theory.’ My first reaction was that this was silly, there are no alternatives…

and his attitude doesn’t seem to have changed since. He and others never discuss the possibility that string theory is simply a wrong idea, a possibility for which the landscape provides overwhelming evidence. This seems to be something he is unwilling to seriously consider.

There’s one peculiar reference in his paper. When he refers to the problem that the landscape can’t even predict the one thing people originally hoped it would be able to, the scale of supersymmetry breaking, he writes:

An obvious question is whether we can understand the supersymmetry-breaking scale (see [68] and references therein). Is low energy supersymmetry, or some alternative [69, 70], favored?

[69] is a reference to split supersymmetry, and [70] is a reference to a paper by Fox et. al called Supersplit Supersymmetry. The strange thing about the apparently serious reference to [70] is that the paper in question is actually an April Fool’s joke (check the date on it…). The authors were making fun of how supersymmetric phenomenology is being pursued, and specifically of the idea of split supersymmetry. In “supersplit supersymmetry”, all superpartners are pushed up to unobservability at the Planck scale, solving all the problems caused by the lack of observation of effects of supersymmetry. For any conceivable purpose, the supersplit supersymmetry model is precisely the standard model. The joke is that particle physicists have become so enamored of supersymmetry that they would happily study a supersymmetric model inherently indistinguishable from the much, much simpler standard model. Part of being unwilling to consider the idea that superstring theory might be wrong is being unwilling to consider the idea that supersymmetry might be wrong, and thus, instead of referring to the standard model, one adopts the supersplit supersymmetry model.

The funny thing about this April Fool’s joke paper is that, according to SPIRES, Polchinski’s is the sixth paper to cite it. Looking through these six papers, only one of them seems to be aware of the joke, with a footnote to the reference pointing out that the paper appeared on April Fool’s day and that the model was equivalent to the standard model. Some of the particle theory community now seems to think that the idea of a vastly more complicated model that is inherently indistinguishable from the standard model but is in some sense “supersymmetric” is a model worth taking seriously and making reference to. Yet another weird thing about this paper is that there is one trackback to it, and this trackback was generated by a comment from LambchopofGod on a Cosmic Variance posting. Presumably it’s a bug, not a feature, that any commenter on an approved blog can generate trackbacks at the arXiv, but seeing the way the arXiv handles trackbacks, who knows. In any case we’ll see at some point if trackbacks to the Polchinski paper and the Fox et. al. paper get generated from this posting. Hopefully at least a trackback to the Fox et. al. paper will appear, since many readers of that paper don’t seem to realize that


This entry was posted in Uncategorized. Bookmark the permalink.

63 Responses to Polchinski on the Landscape

  1. Aaron Bergman says:

    To the point on the analogy between the anthropic principle and natural selection, it might be an oomphless tautology, but it doesn’t need the same oomph as natural selection. As long as there’s one universe with life in it, we could be in that universe.

  2. woit says:


    It’s also true that we could all be a figment in the imagination of the Flying Spaghetti Monster. I wish someone out there who doesn’t think this stuff is indefensible pseudo-science would, instead of defending it by invoking tautologies, give a plausible scenario of how it will lead to the kind of testable predictions that up until now everyone agreed were required before you called something science.

  3. a philosophy student says:


    My point is that Polchinski, besides like Peter says at other places in his paper, in this quote mixes the resp. situations of evolution and anthropics by his chosing of terms.

    AP is a tautology, and as such, states facts. What Peter would like physicists to do is to search for a reason for these facts, a mechanism that explains the state AP refers to.

    Now NS is such a mechanism, but of course for a different state of facts, namely for what I above, for easy of the discussion, have called NP.

    AP does not have such a mechanism, and it is wrong to put both AP and NP on equal footing, as Polchinski does in the above quote, making it appear they are.

    He should have written, instead of ‘natural selection’, something like ’survivors survive because they càn survive’ (NP). This is a statement of facts, which can be, liguistically, and more important: logically, a tautology, but the term ‘natural selection’ cannot be a tautology for it is an empirical mechanism, nót a statement of facts which can be a logical tautology. But of course doing so would point out AP has no explanatory mechanism.

    The correct use of these terms in this is important in (@ least) two respects:

    1) in the context of string theory: as Peter points out:

    … trying to blur the difference between these two very different situations is not something a scientist should be doing …

    …… physicists who care about this credibility (of science) should not
    be claiming that an idea like the anthropic landscape has the same status as heavily tested and verified theories like that of evolution.

    2) these days Darwin is under attack from ‘unscientific’ corners. It is in this respect that it is very important to be clear in one’s choice of words. You cannot describe natural selection by ‘survivors survive’. NS is much much more than that. By bringing NS (wrongfully) down to the level of a statement of facts it can be eassily misunderstood and attacked, eg by calling it a tautology.

    On Scott’s post:

    … natural selection is an “algorithmic workhorse” — a mechanism for amplifying low-probability events to higher probability. Start it up, and it takes you the rest of the way

    This makes it appear as if NS is an always-sure way to an always-sure outcome. NS is an algorithm, granted, but one that relies heavily, not to say totally, on external input, ie the environment. This is a highly coincidential matter. Take eg the extinction of the dinos. Whether it was by an asteroid or not, fact is had they lived on, mammals would still be small mouse-like creatures hiding in small holes in the ground.

    So, NS also has ‘an sich’ nothing to do with the probability of events. It will not change the low or high-probability of an event. It’s all coincidence. It is a fact that human life is a very low-probability event in the universe, but that it excists here is not,
    qua probability, due to NS, but to the occurence of the low-probability fact that after earth was formed, it gave rise to the necessary ‘materials’ for evolving life. This ‘material’ is then worked upon by the evolutionary mechanism NS, but the latter has nothing to do with the probability of the correct circumstances for evolving life being present.

    Note: of course it’s not only the circumstances on earth that are relevant, or low-probable, but also the distance earth-sun, the presence of a big object like jupiter swallowing comets and asteroids, the distance sun-center of the galaxy, aso.

    A note on Jupiter: the point of Jupiter swallowing collision-course-objects means that life on earth has got the necessary time to evolve, as the dinos will be happy to point out.

    Another note: this whole discussion depends on how you weigh the term ‘low-probability’. 1/zillion may be considered low probable, but in an infinite set this still gives rise to an infinite amount. According to this reasoning, if our universe, and/or the multiverse, and/or whatever is infinite, then there are an infinite of worlds with intelligent human-like things on it – even discussing the same things we do now.

    Even by this account it’s not due to NS that there is a sure outcome of whatever you want, but it is due to the infinity of the ‘initial state’ it works upon.

    So NS doesn’t in fact necessarily has ‘oomph’ – @ least not when not considering the ‘basic block’, ie the gene, or, in Smolin’s case, the black hole – and even in this case, again, it’s not due to NS but to the
    nature of the material it works upon.


    Sorry Peter if this is way to off, I only meant to make a humble first post here.

  4. The Lithic Principle. The Liquid Principle. The Canid Principle (of course, man´s best friend must have some importance on all this). In summary, a multitude of principles to chose from in order to study the multiverse. I am sure among them there must be a Multiverse Principle (the Universe is the way it is so to allow for the existence of a multitude of others).

    Now, seriously, I am deeply intrigued that I exist and see the Universe around me. I really would like to know why things are like that. The Anthropic Principle is fruitless to that end, in my opinion, by its own definition.


  5. Aaron Bergman says:

    Peter, we really don’t have to go through this yet one more time.

    Cosmological natural selection doesn’t have the “oomph” that the real thing does, either, because nothing ever dies.

  6. Who says:

    *because nothing ever dies*

    entire universes die

    Aaron you brought up CNS. Scott did not mention it, nor did I. Is this because you have some new ideas about it, like “nothing ever dies”, and want to discuss them?

  7. Aaron Bergman says:

    One can read between the lines in your comments.

    The point is that one does not need a way to populate the landscape, a way for “nice universes” to reproduce or anything. Once you have a mechanism that produces a single universe which can support life, the anthropic principle holds.

    There are a zillion different ways one can put some sort of measure on the space of vacua, if you feel like engaging in some sort of principle of mediocrity, but that’s not the anthropic principle.

  8. Who says:

    You already said that Aaron, and Peter rebutted it. You said
    *… As long as there’s one universe with life in it, we could be in that universe.*

    Peter replied*…give a plausible scenario of how it will lead to the kind of testable predictions…required before you {call] something science*

    Scott’s original point was:
    *…natural selection is an “algorithmic workhorse” — a mechanism for amplifying low-probability events to higher probability. Start it up, and it takes you the rest of the way…*

    It is a powerful means of generating complex organization. (Perhaps the most effective mechanism we know of for doing this.)
    Mere happenstance is not. Therefore, I think Scott’s point was, Polchinski made a poor analogy.

    Do you think Polchinski made a correct analogy? So far your comment does not support this, but has been to repeat that a natural mechanism for self-tuning or amplifying low-probability events is NOT NEEDED, which is not quite on the mark. If you would like, please think up a reason why Polchinski’s comparison of biological Natural Selection to string thinkers’ Landscape speculation should not be considered far-fetched.

  9. Aaron Bergman says:

    Peter’s statement, true or not, is not a rebuttal.

    I don’t really care all that much about how good the analogy between the anthropic principle and natural selection is; it’s obviously not a precise analogy. I’m just trying to clarify a few issues along the way.

  10. ksh95 says:

    Peter Said:

    …”give a plausible scenario of how it will lead to the kind of testable predictions that up until now everyone agreed were required before you called something science…”

    Didn’t Smolin do precisely this? All one needs is for the “Universe-Creating-Mechinism” to be dependent on a few measurable parameters.

  11. Aaron Bergman says:

    It depends on what you mean by predictive. There are a variety of way to populate a landscape of vacua; common examples are some sort of quantum mechanical superposition, tunneling, Smolin’s idea, etc. These will give different distributions of the population of vacua. The fundamental leap, however, is whether this population translates into a probability measure. If you make that leap, a sort of principle of mediocrity, you can make probabilistic predictions. Otherwise, one is simply left with the binary question of whether a particular vacuum gets populated or not.

  12. Chris W. says:

    The most galling thing to me about drawing the analogy with natural selection is that the variation upon which selection operates in biology is an inescapable empirical fact, whether or not you understand the underlying mechanism (and Darwin didn’t). This observed variation was Darwin’s logical starting point. He knew that it was the raw material with which animal and plant breeders worked, and realized that the essential logic of selective breeding applied even when the relevant contingencies—selective conditions—were not consciously arranged but arose accidentally, or as a result of the internal dynamics of an ecosystem.

    In contrast, the multiverse is a purely theoretical construct. It is taken seriously mainly because it seems to follow from the joint application of quantum field theory and general relativity in cosmology. However, even this line of argument seems to have degenerated, insofar as the putative existence of the multiverse is now being taken to imply that observed laws in our “pocket” of the multiverse are “frozen accidents”, perhaps anthropically selected. This strikes me as self-defeating to the point of silliness, inasmuch as it was the ramifications of those laws that led us to assert the existence of the multiverse in first place.

    (Lately I’ve been wondering why these points aren’t made more often. The self-undermining implications of the multiverse idea have occasionally been pointed out by other commenters on this blog.)

  13. Pingback: Not Even Wrong » Blog Archive » arXiv Weirdness

Comments are closed.