We’ve returned from the lovely city of New Orleans, where within a short period of time I was able to sample shrimp and grits, bread pudding soufflé, turtle soup, chicken gumbo, soft-shelled crab with crawfish étouffée, and of course beignets. Oh yes, also participated in the Greer-Heard Forum, where I debated William Lane Craig, and then continued the discussion the next day along with Alex Rosenberg, Tim Maudlin, James Sinclair, and Robin Collins. The whole event was recorded, and will be released on the internet soon — hopefully within a couple of days.
[Update: Here is the video.]
In the meantime I thought I’d provide some quick post-debate reflections. Overall I think it went pretty well, although I certainly could have done better. Then again I’m biased, both by being hard on myself in terms of the debate performance, but understandably of the opinion that my actual ideas are correct. I think I mostly reached my primary goal of explaining why many of us think theism is undermined by modern science, and in particular why there is no support to be found for it in modern cosmology. For other perspectives see Rational Skepticism or the Reasonable Faith forums.
Short version: I think it went well, although I can easily think of several ways I could have done better. On the substance, my major points were that the demand for “causes” and “explanations” is completely inappropriate for modern fundamental physics/cosmology, and that theism is not taken seriously in professional cosmological circles because it is hopelessly ill-defined (no matter what happens in the universe, you can argue that God would have wanted it that way). He defended two of his favorite arguments, the “cosmological argument” and the fine-tuning argument; no real surprises there. In terms of style, from my perspective things got a bit frustrating, because the following pattern repeated multiple times: Craig would make an argument, I would reply, and Craig would just repeat the original argument. For example, he said that Boltzmann Brains were a problem for the multiverse; I said that they were a problem for certain multiverse models but not others, which is actually good because they help us to distinguish viable from non-viable models; and his response was the multiverse was not a viable theory because of the Boltzmann Brain problem. Or, he said that if the universe began to exist there must be a transcendent cause; I said that everyday notions of causation don’t apply to the beginning of the universe and explained why they might apply approximately inside the universe but not to it; and his response was that if the universe could just pop into existence, why not bicycles? I was honestly a bit surprised at the lack of real-time interaction, since one of Craig’s supporters’ biggest complaints is that his opponents don’t ever directly respond to his points, and I tried hard to do exactly that. To be fair, I bypassed some of his arguments (see below) because I thought they were irrelevant, and wanted to focus on the important issues; he might feel differently. I’m sure that others will have their own opinions, but soon enough the videos will allow all to judge for themselves. Overall I was moderately satisfied that I made the responses I had hoped to make, clarified some points, and gave folks something to think about.
Longer version (much longer, sorry): the format was 20-minute opening talks by each speaker (Craig going first), followed by 12-minute rebuttals, and then 8-minute closing statements. Among the pre-debate advice I was given was “make it a discussion, not a debate” and “don’t let WLC speak first,” both of which I intentionally ignored. I wanted all along to play by his rules, in front of his crowd, and do the best job I could do without any excuses.
In his opening speech Craig gave two arguments: the Kalam Cosmological Argument (the universe must be caused, and the cause is God), and the teleological/fine-tuning argument (the parameters of the universe appear designed for the existence of life). To his great credit, WLC actually stuck to arguments concerning physical cosmology, where presumably my expertise would be most valuable; he didn’t hide behind primarily metaphysical arguments like the ontological argument or the denial of realized infinities. The two he used were familiar from his repertoire, and they were the two that I was primarily interested in talking about myself, so we were off. (I’ll try to reconstruct the logic rather than doing a point-by-point recap, since I’m mostly working from memory. Naturally, my memory of my own parts will be sharper than my memory of WLC’s, so I’ll happily accept factual corrections.)
The cosmological argument has two premises: (1) If the universe had a beginning, it has a transcendent cause; and (2) The universe had a beginning. He took (1) as perfectly obvious, and put his effort into establishing (2). Partly he used the celebrated (by theologians) Borde-Guth-Vilenkin theorem, which says that a universe with an average expansion rate greater than zero must be geodesically incomplete in the past. But he also used an argument I hadn’t heard before: from the Second Law of Thermodynamics (entropy in a closed system doesn’t decrease). I think the argument was basically that the Second Law implies that we approach equilibrium, and in an infinitely-old universe we should therefore have reached equilibrium long ago, which we haven’t, so the universe began at some finite time in the past.
My attitude toward the above two premises is that (2) is completely uncertain, while the “obvious” one (1) is flat-out false. Or not even false, as I put it, because the notion of a “cause” isn’t part of an appropriate vocabulary to use for discussing fundamental physics. Rather, modern physical models take the form of unbreakable patterns — laws of Nature — that persist without any external causes. The Aristotelian analysis of causes is outdated when it comes to modern fundamental physics; what matters is whether you can find a formal mathematical model that accounts for the data. The Hartle-Hawking “no-boundary proposal” for the wave function of the universe, for example, is completely self-contained, not requiring any external cause.
Mostly Craig ignored this argument, which to me was the most important part of the debate. In the first rebuttal he said that the Hartle-Hawking model was indeed lacking something — a reason why the universe exists at all. To me this looks like confusing the cosmological argument with the argument from contingency, but since my objection applied to that case as well I didn’t raise that as an rebuttal. Rather, I pointed out that this response sailed right by my actual argument, which was that a self-contained physical model is all you need, and asking for anything more is completely unwarranted. To drive the point home, I elaborated on why things like “causes” and “explanations” make perfect sense for parts of the universe, but not for the universe itself: namely, that we live in a world with unbreakable patterns (laws of physics) and an arrow of time, but the universe itself (or the multiverse) is not one element of a much bigger pattern, it’s all there is. Finally in the closing speech WLC finally offered arguments in favor of the idea that the beginning of the universe implies a transcendental cause: (1) it’s a metaphysical principle; (2) if universe could pop into existence, why not bicycles?; and (3) there’s no reason to treat the universe differently than things inside the universe. To me, (1) isn’t actually an argument, just a restatement; and I had already explained why (2) and (3) were not true, and he didn’t actually respond to my explanation. So by the time my rebuttal came around I didn’t have much more new to say. Craig spent some time mocking the very idea that the universe could just “pop into existence.” I explained that this isn’t the right way to think about these models, which are better understood as “the universe has an earliest moment of time,” which doesn’t misleadingly appeal to our intuitions of temporal sequence; but my explanation seemed to have no effect.
The second premise of the Kalam argument is that the universe began to exist. Which may even be true! But we certainly don’t know, or even have strong reasons to think one way or the other. Craig thinks we do have a strong reason, the Borde-Guth-Vilenkin theorem. So I explained what every physicist who has thought about the issue understands: that the real world is governed by quantum mechanics, and the BGV theorem assumes a classical spacetime, so it says nothing definitive about what actually happens in the universe; it is only a guideline to when our classical description breaks down. Indeed, I quoted a stronger theorem, the “Quantum Eternity Theorem” (QET) — under conventional quantum mechanics, any universe with a non-zero energy and a time-independent Hamiltonian will necessarily last forever toward both the past and the future. For convenience I quoted my own paper as a reference, although I’m surely not the first to figure it out; it’s a fairly trivial result once you think about it. (The Hartle-Hawking model is not eternal to the past, which is fine because they imagine a universe with zero energy. In that situation time is an approximation rather than fundamental in any case — that’s the “problem of time” in quantum gravity.)
Sadly, Craig never responded to my point about the QET. Instead, he emphasized another “theorem” in a paper by Aron Wall. This is a great paper, well worth reading — but it doesn’t say what Craig wants it to say, which I was only able to check after the debate. Wall (like BGV) proves theorems that apply to semiclassical gravity (classical spacetime with propagating quantum fields — see comment from Aron below), and then speculates “the results may hold in full quantum gravity” and “there is a reasonable possibility that the Penrose singularity theorem can be proven even in the context of full quantum gravity.” As good as the paper is, proving a theorem in the semiclassical case and then opining that it is probably extendable to the full quantum gravity case does not actually represent a “theorem” about the quantum case. And in fact I think it’s highly unlikely to be extendable in the sense Craig wants it to be, since the QET says that’s impossible (unless the universe has zero energy or a time-dependent Hamiltonian, in which case it’s easy to avoid eternity). But I had never seen Wall’s paper before, and Craig didn’t give a precise statement of the purported theorem, only the above quote about “reasonable possibility”; as a result I didn’t know the range of applicability of the “theorem” or its assumptions, so chose not to talk about it rather than making guesses. That was probably a strategic mistake on my part.
While I’m lingering over my mistakes, I made a related one, when Craig emphasized a recent paper by Anthony Aguirre and John Kehayias. They examined the “emergent universe” scenario of George Ellis and Roy Maartens, in which the universe is in a quasi-static pre-Big-Bang state infinitely far into the past. Aguirre and Kehayias showed that such behavior is unstable; you can’t last in a quasi-static state for half of eternity and then start evolving. Personally, I didn’t think this was worth talking about; I completely agree that it’s unstable, I never promoted or defended that particular model, and I just didn’t see the relevance. But he kept bringing it up. Only after the debate did it dawn on me that he takes the specific behavior of that model as representative of any model that has a quantum-gravity regime (the easiest way out of the “beginning” supposedly predicted by the BGV theorem). That’s completely false. Most models with a quantum-gravity phase are nothing like the emergent universe; typically the quantum part of the evolution is temporary, and is surrounded on both sides by classical spacetime. But that’s so false that I didn’t even pick up that WLC was presuming it, so I never responded. Bad debater.
The other argument from Craig in favor of the universe having a beginning comes from the fact that entropy is increasing, in accordance with the Second Law. This is another case where it took me a while to understand the point he was trying to get at. To me, it is perfectly obvious and well-understood that the Second Law comes about because of the configuration of matter in our local region of universe, not because of some ironclad fundamental law. (Otherwise Ludwig Boltzmann lived in vain — and I have his tombstone right up there on my blog header.) A theory like my model with Jennifer Chen tries to explain how the Second Law emerges in local regions of the universe, by showing how a universe with no equilibrium state can evolve forever (rather than settling down), and entropy will naturally increase both in the far past and the far future. Craig seems to think that the Second Law should be absolute, so that models like ours are ruled out because entropy doesn’t increase monotonically — i.e. they “violate” the Second Law. (Unless I’m still misunderstanding his point — his presentation was uncharacteristically muddled here.) This is a pretty straightforward misunderstanding of the origin of the Second Law and the point of our model, although to be fair I caught on too late to present a strong counterargument.
It was interesting that Craig spent so much time talking about the Carroll-Chen model, when I wouldn’t have brought it up at all if left to my own devices. I think the model is very useful as an illustration of an interesting fact: you can derive a natural dynamical origin of the Second Law in a universe that creates new entropy without bound by creating regions of space that look like our Big Bang. But I’m the first to admit that it’s speculative, and especially that the process of baby-universe creation is ill-understood, if it’s possible at all. So I probably wouldn’t have dwelt on it, but Craig really went to town on the model. Unfortunately, nearly everything he said about it was just wrong. First, he tried to claim that having a moment of time in the history of the universe when entropy was lowest counts as a “thermodynamic beginning,” even if there is more universe in both direction of time around that moment. That’s quite an innovative definition (to be polite), but more importantly that kind of “beginning” has nothing to do with the kind of “beginning” where God would create the universe. I made this point, but it wasn’t answered. Relatedly, he seemed to think it was a glaring mistake (or perhaps intentionally subterfuge?) that on one picture of our model I had the “time” axis have only one arrow, while on another version of the picture I put arrows pointing in both directions. I apologize for being sloppy, but it’s neither a mistake nor a dastardly plot; either version is acceptable, because you have a time coordinate that runs monotonically from -infinity to +infinity, but the direction of entropy increase (that defines the arrow of time) is not monotonic. Next, he tried to claim that our model violated unitarity (conservation of quantum information), which is flatly wrong. He supplied two pieces of evidence, in the form of quotes from Stephen Hawking and Chris Weaver. But the Hawking quote was completely out of context; he was talking about the fact that he no longer thought that wormholes would lead to violation of unitarity in black-hole evaporation, nothing to do with cosmology. And the Weaver quote that Craig read had nothing to do with unitarity at all; it merely pointed out that the process of baby-universe creation is speculative and not well-understood, which I’m the first to admit. Again — I didn’t actually hold up this model as a solution to anything, but he felt the need to attack it, so I had to defend its honor a little bit.
In contrast, I wanted to talk about a model developed by Anthony Aguirre and Stephen Gratton. They have a very simple and physically transparent model that (unlike my theory with Chen) imposes a low-entropy boundary condition at a mid-universe “bounce.” It’s a straightforward example of a perfectly well-defined theory that is clearly eternal, one that doesn’t have a beginning, and does so without invoking any hand-waving about quantum gravity. I challenged Craig to explain why this wasn’t a sensible example of an eternal universe, one that was in perfect accord with the BGV theorem, but he didn’t respond. It wasn’t until the talks on the following day that Craig’s teammate James Sinclair admitted that it seemed like a perfectly good model to him.
But again — my main point was not to push this or that specific model, but to argue that it’s the models that matter, not some general theorem in a regime we don’t pretend to understand. So I listed a bunch of plausible-looking eternal cosmologies. The point is not that all or some of these models is perfect; it’s that they’re eternal. So we should judge them on their merits, rather than claiming to have general arguments that there are no such things. (It’s as if WLC has a powerful general theorem against heavier-than-air flying machines, while airplanes keep buzzing overhead.) On occasion it would be as if Craig admitted that there were indeed eternal cosmologies, but they were all ruled out for various unspecified other problems. That sounds like a suspiciously far-reaching claim, but one that is hard to directly dispute without any details being presented. More importantly, it’s beside the point. Which, in case it isn’t yet clear, is that it’s the models that matter, not any general theorems. If there are some cosmological models that are eternal but have other problems, there’s no reason to stop looking for other models that are also eternal but don’t have those problems. This is a case where working scientists are quick to admit that we don’t know the answer, so we shouldn’t stop considering all legitimate possibilities.
The fine-tuning argument proceeded in a more straightforward way. Unlike the cosmological argument, where Craig presented a few twists I hadn’t heard before (though he may have used them in other debates, I don’t know), the fine-tuning presentation was pretty standard. I acknowledged that, unlike the cosmological argument that is based on outdated metaphysics, the fine-tuning argument is a respectable scientific claim: two models trying to account for some data. But I gave five reasons why it was nevertheless not a good argument for theism:
- We don’t really know that the universe is tuned specifically for life, since we don’t know the conditions under which life is possible.
- Fine-tuning for life would only potentially be relevant if we already accepted naturalism; God could create life under arbitrary physical conditions.
- Apparent fine-tunings may be explained by dynamical mechanisms or improved notions of probability.
- The multiverse is a perfectly viable naturalistic explanation.
- If God had finely-tuned the universe for life, it would look very different indeed.
Craig didn’t respond to 2. or 3. here. To counter 1. he simply noted that other physicists disagreed with me, which again really isn’t an argument; he didn’t offer any suggestion that we actually do know the conditions under which life can and cannot form. Against 5. he invoked an argument by Robin Collins that the universe is optimized for “discoverability,” at least when we consider the known physical parameters. To me this argument is completely implausible right on the face of it, since it’s trivially easy to imagine ways to make it easier to discover the universe (just make the Higgs boson lighter!). But I knew Collins was going to give a full discussion of that argument the next day, so I saved my response until then.
In my first speech I used 5. above as a launching pad to make a bigger point: the real reason theism isn’t taken seriously is because it’s completely ill-defined. If we would presume to contemplate theism from an intellectually honest perspective, we would try to decide what kind of universe we would expect to live in if theism were true; then we would do the same for naturalism; and finally we would compare those expectations to the real world. But when we do that we find theistic expectations failing to match reality over and over again. Now, I know perfectly well (from experience as well as from cogitation) that you can never make headway with theists by claiming “If God existed, He would do X, and He doesn’t” (where X is “prevent needless suffering,” “make His existence obvious,” “reveal useful non-trivial information to us,” “spread religious messages uniformly over the world,” etc.) Because they have always thought through these, and can come up with an explanation why God would never have done that. (According to Alvin Plantinga, our world — you know, the one with the Black Death, the Holocaust, AIDS, Hurricane Katrina, and so on — is “so good that no world could be appreciably better.”) But these apologetic moves come at a price: they imply a notion of theism so flexible that it becomes completely ill-defined. That’s the real problem. Craig’s way of putting it is to suggest that God is “like the cosmic artist who wants to splash his canvas with extravagance of design.” That’s precisely why naturalism has pulled so far ahead of theism in the intellectual race to best model our world: because it plays by rules and provides real explanations for why the world is this way rather than that way.
Against the multiverse, Craig’s major argument (surprisingly) was the Boltzmann Brain problem. I say “surprisingly” because it’s such an easy argument to rebut. Sure, Boltzmann Brains are a problem — for those models with a Boltzmann Brain problem. Not all models have them! And a good modern multiverse cosmologist focuses on those models that avoid them. In this sense, the BB problem is a good thing; it helps us distinguish viable models from non-viable ones. As far as I can tell, this straightforward response was completely ignored by Craig. He just kept repeating that Boltzmann Brains were really bad things. He aimed this criticism particularly at the Carroll-Chen model, which I would say is very bad aim; it’s much less likely that BB’s are a problem in our scenario than in most other multiverse theories, since you actually produce baby universes (with potentially billions of observers) more frequently than you produce individual Boltzmann Brains. But I didn’t emphasize that point, since my goal wasn’t to defend that particular model.
As an aside, the Boltzmann Brain discussion illustrated a problem with the debate as a whole: it was too technical (and I think mostly on WLC’s side, although I deserved some of the blame). I had prepared a single slide about the topic of BB’s, but upon reflection I figured that many people wouldn’t be familiar with the term, so I replaced it with “observers as random fluctuations.” But Craig just plowed right into the technicalities, without trying to explain what the Boltzmann Brain problem really was. When my term came I gave a one-sentence definition, but in retrospect that wasn’t really sufficient, so on Saturday (where we both gave five-minute responses to a series of 40-minute talks by the four other speakers) I took a few minutes to give a more careful explanation. It’s probably an inherent flaw of the format; if a speaker takes a minute to do a bit of pedagogy rather than argumentation, the audience benefits but the speaker suffers. So the game theory tells you to be obscure, to the general detriment.
A couple of words about the rebuttal speeches in particular. Each of us pulled one “gotcha” move in the rebuttals; mine worked, his didn’t (in my opinion, of course). His referred to fine-tuning: he found a quote by me saying that the low entropy of the early universe seems finely-tuned. Which is true! But he claimed that I had said we didn’t know whether fine-tuning was real. That part is completely false. I had said we didn’t know that fine-tuning for life was real; indeed, the words “for life” were italicized on my slide. And it’s perfectly obvious that the fine-tuning of our initial entropy isn’t necessary for life; the entropy could have been fantastically larger than it actually was and life still could have arisen. (For example, you could imagine a universe with our Solar System as it essentially is, but otherwise in thermal equilibrium.)
On my part, I knew that WLC liked to glide from the BGV theorem (which says that classical spacetime description fails in the past) to the stronger statement that the universe probably had a beginning, even though the latter is not implied by the former. And his favorite weapon is to use quotes from Alex Vilenkin, one of the authors of the BGV theorem. So I talked to Alan Guth, and he was gracious enough to agree to let me take pictures of him holding up signs with his perspective: namely, that the universe probably didn’t have a beginning, and is very likely eternal. Now, why would an author of the BGV theorem say such a thing? For exactly the reasons I was giving all along: the theorem says nothing definitive about the real universe, it is only a constraint on the classical regime. What matters are models, not theorems, and different scientists will quite naturally have different opinions about which types of models are most likely to prove fruitful once we understand things better. In Vilenkin’s opinion, the best models (in terms of being well-defined and accounting for the data) are ones with a beginning. In Guth’s opinion, the best models are ones that are eternal. And they are welcome to disagree, because we don’t know the answer! Not knowing the answer is perfectly fine. What’s not fine is pretending that we do know the answer, and using that pretend-knowledge to draw premature theological conclusions. (Chatter on Twitter reveals theists scrambling to find previous examples of Guth saying the universe probably had a beginning. As far as I can tell Alan was there talking about inflation beginning, not the universe, which is completely different. But it doesn’t matter; good scientists, it turns out, will actually change their minds in response to thinking about things.)
I very much hope that I hammered these points home enough to help clarify issues in the minds of listeners/readers. But from Craig’s (lack of) reaction, and from the online discussion from his supporters, I doubt it will make any difference. He will continue to quote Vilenkin saying the universe probably had a beginning, which is fine because that’s what Vilenkin actually thinks. He will not start adding in the fact that Guth thinks the universe is probably eternal, nor will he take the even more respectable position of not relying on people’s individual opinions at all and simply admitting that we don’t have good scientific reasons to think one way or the other at the moment. But we’ll see. (And to reiterate: I think the whole discussion is enormously less important than the bigger point that a “cause” is completely unnecessary even if the universe did have a beginning.)
For my closing statement, I couldn’t think of many responses to Craig’s closing statement that wouldn’t have simply be me reiterating points from my first two speeches. So I took the opportunity to pull back a little and look at the bigger picture. Namely: we’re talking about “God and Cosmology,” but nobody really becomes a believer in God because it provides the best cosmology. They become theists for other reasons, and the cosmology comes later. That’s because religion is enormously more than theism. Most people become religious for other (non-epistemic) reasons: it provides meaning and purpose, or a sense of community, or a way to be in contact with something transcendent, or simply because it’s an important part of their culture. The problem is that theism, while not identical to religion, forms its basis, at least in most Western religions. So — maybe, I suggested, tentatively — that could change. I give theists a hard time for not accepting the implications of modern science, but I am also happy to give naturalists a hard time when they don’t appreciate the enormous task we face in answering all of the questions that we used to think were answered by God. We don’t have final answers to the deep questions of meaning and fulfillment and what it means to lead a good life. Religion doesn’t have the final answers, either; but maybe it has learned something interesting over the course of thousands of years of thinking about these issues. Maybe there is some wisdom to be mined from religious traditions, even for naturalists (which everyone should be).
More than once of the course of the weekend I spoke conciliatory-sounding words about how we’re really all in this together, theists and naturalists, trying to understand the deep questions in a confusing world. And I meant all of it, in complete sincerity. I will be absolutely uncompromising about what I think the truth is concerning questions of substance; but I don’t ever want to start thinking of people who disagree with me about those questions as my enemies. Many times in New Orleans, people on “the other side” came up after my presentations (in which I said that their most deeply held beliefs had been definitively refuted by the progress of modern science) and thanked me, saying how much they appreciated my ideas and that I had given them something to think about, even if they remained quite resolute in their beliefs. Not every single person was so gracious, but the vast majority. I admire those folks, and I hope I can be as sincere and open when people who I disagree with are speaking in good faith.
To me, Craig’s best moment of the weekend came at the very end, as part of the summary panel discussion. Earlier in the day, Tim Maudlin (who gave an great pro-naturalism talk, explaining that God’s existence wouldn’t have any moral consequences even if it were true) had grumped a little bit about the format. His point was that formal point-counterpoint debates aren’t really the way philosophy is done, which would be closer to a Socratic discussion where issues can be clarified and extended more efficiently. And I agree with that, as far as it goes. But Craig had a robust response, which I also agree with: yes, a debate like this isn’t how philosophy is done, but there are things worth doing other than philosophy, or even teaching philosophy. He said, candidly, that the advantage of the debate format is that it brings out audiences, who find a bit of give-and-take more exciting than a lecture or series of lectures. It’s hard to teach subtle and tricky concepts in such a format, but that’s always a hard thing to do; the point is that if you get the audience there in the first place, a good debater can at least plant a few new ideas in their heads, and hopefully inspire them to take the initiative and learn more on their own.
I completely agree. Events like this are valuable, not because they are efficient ways to find the truth, nor even because there is any reasonable chance of changing the minds of people who are relatively secure in their beliefs (on either side). It’s because there are a lot of people who are not secure in their beliefs, or at least are curious and willing to listen to a variety of ideas. If we think we have good ideas, we should do everything we can to bring them to as many people as possible. I think science and naturalism include some pretty awesome ideas, and I’m happy to share them with as many different people as I can.
p.s. Sorry I didn’t talk about Saturday’s talks, in which Tim Maudlin discussed the relationship between theism and morality (claiming there isn’t any), Alex Rosenberg drew connections between thermodynamics and natural selection to argue that theism is incompatible with Darwinism, Robin Collins argued that the discoverability of the universe is evidence for theism, and James Sinclair talked about notions of time and the origin of the universe. But I’m kind of all talked out on this topic for right now.