Episode Transcript
[00:00:00] Speaker A: There's a methodological principle. A theory that predicts a result outranks the one that merely accommodates and absorbs it, right? And so a theory is tested by what it explains and what it predicts.
And I would say that just as a matter of general epistemology and metaphysics, that theism has a leg up there over naturalism from the get go.
ID the Future, a podcast about evolution and Intelligent design.
[00:00:33] Speaker B: You're likely familiar with the evidence for the fine tuning of the universe, and you may be aware that there are several cosmological models attempting to explain away that fine tuning, including the popular multiverse hypothesis. But did you know there's an actual method for grading these models? Welcome to Idea the Future. I'm your host Andrew McDermott. Today I begin a conversation about evaluating cosmological models with philosopher of physics Bruce Gordon and physicist Brian Miller.
Bruce is Associate Research Director and a Senior Fellow with Discovery Institute's center for Science and Culture. He's also a Research professor of Philosophy of Science at St. Constantine College. Bruce is an expert in cosmology, general relativity, and quantum mechanics. He received his Ph.D. in the history and philosophy of physics from Northwestern University in Chicago, and he holds master's degrees in analytic Philosophy from the University of Calgary and Systematic Theology from Westminster Theological Seminary in Philadelphia. Bruce is the contributing editor of two books, the Nature of Nature and Biological New Perspectives, and is the author of numerous articles and essays and journals and academic volumes including Minding the Models of the Mind, Information and Empirical Science that one published by Discovery Institute Press in recent years. Now Brian is a Senior Fellow and Research Coordinator for the center for Science and Culture here at Discovery Institute Institute. He helps manage the ID 3.0 research program and helped launch the biennial Conference on Engineering and Living Systems, or Cells. He obtained a B.S. in Physics with a minor in Engineering from MIT and a Ph.D. in Complex Systems Physics from Duke University.
His research focuses on thermodynamics, information theory, protein rarity, and the origin of life. He has contributed to multiple books and technical journals covering the debate over intelligent design, including the Mystery of Life's Origin, the Continuing Controversy, the Comprehensive Guide to Science, and Faith and Inference Review Gentlemen, welcome to the show.
[00:02:40] Speaker C: It's a pleasure to be here.
[00:02:41] Speaker A: Pleasure to be here, yes, and it's
[00:02:43] Speaker B: a pleasure to be in such esteemed company today. Now, Brian, you and I had a preliminary discussion on the podcast recently about a lively three hour debate between science communicator Phil Halper and Dr. Stephen Meyer.
Now, for those who missed that discussion or the debate itself can you review for us who Phil Halper is and what the debate covered?
[00:03:05] Speaker C: Yes, certainly. Phil Halper is a British science communicator and popularizer of cosmology.
And he's also a fellow of the Royal Astronomical Society. And he's best known for running the podcast before the Big Bang, where he has featured numerous prominent cosmologists talking about the origin of the universe and the philosophical implications of it.
And in the debate with Phil Helper, what Phil Halper had with Steve Meyer is they addressed the arguments from Stephen Meyer's book Return of the God Hypothesis, related to the beginning of the universe and the fine tuning of the universe, both pointing to the universe having a creator. And during the debate, they also talked about the plausibility of a few dozen models that have been proposed that both attempt to remove the beginning of the universe or explain away the fine tuning. And they talked about the plausibility of the models and whether they truly removed the need for a creator.
[00:04:00] Speaker B: Okay, that might explain why it went three hours long. Was it a good exchange or did it get heated at times?
[00:04:08] Speaker C: Well, it was certainly spirited, but it was a very thoughtful exchange. They were very respectful, they presented very strong arguments, and I think it was one of the best debates I've seen in terms of producing genuine understanding in the different positions.
[00:04:23] Speaker B: Okay, Bruce, your opinion of the debate?
[00:04:26] Speaker A: Well, I don't have a lot to add to what Brian has said. I do think that it was exemplary in a lot of ways, and it is a good example of how these debates should go, all things considered. It generated more light than heat. I mean, Phil's a knowledgeable debater.
He was courteous face to face, and he's a very good communicator. And as Brian pointed out, he's had the opportunity to interview most of the leading cosmologists in the world. And of course, he's written the book, what is it?
The Battle of the Big Bang with Naish Afshordi, who's physicist at the University of Waterloo and the Perimeter Institute in Ontario.
So he knows these models. And, um, Steve, on the other hand, as he often is, was superb in the debate. He was erudite, he was on point.
He was able to draw on specific examples when it was appropriate to do so.
And in terms of assessing the whole interaction, I. I think that Brian's comment in the last dialogue he had with you was. Was appropriate in that respect. He said, basically, it comes down to the priors in a Bayesian analysis of things. And maybe by the end of our discussions today, we'll get to see just why that's the case and how extreme the prior has to be in order to wipe out the Bayes factor. That would tip the scales in favor of design, but that's territory yet to be covered. So in any case, I'm hoping that today we might be able to explore a couple of the gaps that were left. I mean, it was a three hour discussion, but there were still some things that they didn't quite get to or didn't get explored in detail.
I'm hoping we might be able to talk about Phil's version of the normalizability objection in terms of his inscrutable dice and Christopher Hitchcock's jar of beans as an example of problems with fine tuning arguments.
And Steve didn't quite get to say everything I think he might have about bounce cosmologies either. So maybe we can get to that and then move on from there to a discussion of how we go about quantitatively ranking these different cosmological explanations.
[00:07:08] Speaker B: Yeah, that's a great plan. Thanks for laying that out for us, Bruce.
Now, Brian, in our previous conversation together you said that Bruce has a method that shows how every single model that attempts to remove a beginning or explains away fine tuning has features that point to a designer. So we're going to unpack that evaluative instrument he calls it or scorecard. But before that, we're going to start by looking at the key questions that the Halper Meier debate turns on. And Brian, as we do that, I'm going to hand the reins to you so you can ask Bruce some relevant questions. I may jump in here and there, but I want to let the physicists have some dialogue here. And of course we want a nice dose of technical detail, but let's do our best to share easy to understand examples and concise explanation too. So Brian, why don't you take over?
[00:07:58] Speaker C: Thank you, Andrew. And it really is a pleasure to interview you, Bruce, because you have done some cutting edge work on this topic and it's really brilliant.
So I want to begin by just talking about Phil strategy during the debate where he basically said there's all these different models that have been proposed that explain away the beginning or they explain away fine tuning. And he said that he wasn't convinced by any of the models, but he fact he but he felt that there was so many models that at least one of them had to be describing what's actually true. Much like if you buy enough winning lot, enough lottery tickets, at least one of those tickets should win.
How do you feel about that general argument, Bruce?
[00:08:47] Speaker A: Well, I don't feel very good about the argument, to tell you the truth.
So, I mean, his point is, yeah, I don't really believe any of these specifically, but we've got a whole raft of them here. I mean, their book Battle of the Big bang talks about 25. I once did a tally and counted over 40 different cosmological models that are in play in discussion, and they're probably more than that.
So is it like a lottery?
And I think the answer is no, it's not really.
It's an appealing analogy, but I mean, with an actual lottery, you've got a defined ticket space, right? You know how many tickets there are and what's on them.
Then there's a mechanism that selects the winning number that matches the ticket. And cosmology doesn't supply any of those things.
There's no closed, well defined list of possible models, and theorists keep inventing new ones and nobody can enumerate that space in advance.
And beyond that, there's no mechanism for saying, hey, this one's true, right?
And you're always going to have the advocates fighting with each other over which model is superior and why. So without a defined ticket space and without a defined drawing that says, yeah, this ticket wins, we're not dealing with an inference at all. We're dealing with an ill defined hope of some sort. That's basically all he's got going for him in this respect. And the models themselves aren't that encouraging, to tell you the truth. I mean, Steve kind of put his finger on it at one point in the debate. He talked about progressive and regressive or degenerating research programs, drawing on the philosopher of science, Simri Lakatosh. So progressive science is going to be making novel predictions, getting confirmed, growing more unified over time, whereas a degenerating research program keeps generating fresh hypotheses whose only function is to try to fill a gap or to protect prior commitments from evidence that seems to be pulling in a different direction. And they're each bought at the cost of new assumptions that are often ad hoc and none of them are independently testable. And if you want a description of what's going on with over 40 models in contemporary cosmology, that's pretty much it. We're looking at what seems to be a degenerative research program, and I think Steve called that quite nicely. We'll maybe get a chance to score these models and see just how degenerative the research program is before we're done here.
[00:11:37] Speaker C: Thank you. That's really helpful. And I think a point Steve has made which was really insightful is that when you look at the history of science, whenever a scientific paradigm gets to a point where it's constantly generating new contradictory models, that suggests the fundamental assumptions are false. So that's a very insightful comment that both you and Steve have made.
Now also an interesting point is how do we even evaluate this entire question?
Because many people in the intelligent design community have argued that the universe has properties that are what we'd expect if there was a designer behind the universe and behind life and behind all of nature. Now let's flip the question. How would you answer the question what would a naturalistic perspective or the view that there is no designer but just the laws of nature, what would that predict about what we observe in the nature? And is that prediction true?
[00:12:32] Speaker A: Well, contrary to a remark that Richard Dawkins made years ago that the universe is exactly what you'd expect if there were no design and everything were random.
I think that we're dealing with a different situation, as you pointed out. I mean, historically speaking, modern science arose out of a theistic worldview in the medieval Christian University, generally speaking, that assumed that the universe was intelligible, that it was orderly, that there was a contingent rational order that would be discernible if we looked.
And so theism makes a bet in that direction. And it's a bet, quite frankly, it could have lost.
But it didn't.
And as a consequence we have modern science now, if we turn it around and we say what does undirected nature positively lead us to expect?
Not what it might be able to accommodate after the fact, but what does it predict in advance?
I think the answer has got to be it doesn't predict much, it's very little blind. Nature doesn't have an end in view, gives us no reason to expect any particular value of a constant over any other value.
It's non informative in that way.
And that's precisely why physicists assign non informative distributions in the fine tuning calculations and discover much to their amazement in many cases just how fine tuned those constants are if you're going to have life.
So standard naturalistic responses, it's a brute fact, no explanation needed, or it's a multiverse. Well, I mean the multiverse is kind of retrofitted to the evidence. The brute fact assertion has problems of its own related to the principle of sufficient reason that we could get into it some point, if you want to talk about those things.
But essentially naturalism is retrofitting it's constructing after the tuning is discovered and trying to explain it away and absorb it rather than expecting it in the first place. Which is precisely what theism does. It's not a surprise for the theist that how does the psalm put it, that the heavens are telling the glory of God.
That was to be expected.
So what would we say then, as a methodological principle, a theory that predicts a result outranks the one that merely accommodates and absorbs it. Right. And so a theory is tested by what it explains and what it predicts.
And I would say that just as a matter of general epistemology and metaphysics, that theism has a leg up there over naturalism from the get go.
[00:15:36] Speaker C: Well, you mentioned this idea that the laws of nature were carefully set or fine tuned for life, but I know that there's been pushback on that argument. For instance, people have talked about what's referred to as the inscrutable dice. Could you explain that response and why it's not valid?
[00:15:53] Speaker A: Well, Holper, in the debate with Steve talked about thinking of it in terms of suppose the constants of physics were set by rolling dice. Right.
And you call a life permitting outcome improbable.
Well, what would you need in order to be able to say that? You need to know how many sides the dice have, whether they're weighted, how many times they were rolled and those sorts of things. And he says we don't know any of that.
So fine tuning probability, he says, isn't merely unknown, it's not well defined.
Okay, so that's the inscrutable dice sort of objection to fine tuning.
And it's a version, and Steve very astutely noted this in the context of the debate of the normalizability objection. Right. That genuine uniform probabilities spread over an infinite range cannot be normalized. You can't make the probabilities add up to one, which they have to be if they're genuine probabilities. Right. So distribute the probability evenly over infinitely many possibilities and every finite slice gets zero.
And that doesn't work. It's incoherent. And if that's what's going on with fine tuning, then there's a significant problem that needs to be addressed.
And that's what Phil was pressing, was a version of this normalizability objection. It's a real objection with a significant pedigree and it deserves good answer.
So one thing we can say is that we don't really need dice with an infinite number of sides. Why not? Well, the argument doesn't require an unbounded range. Physics itself supplies a boundary.
Right at the top we got the Planck scale, such a natural ceiling on energies and field strengths and physical concepts break down beyond it. And at the lower end, we got nuclear and atomic physics constraining the forces and how strong they can be and still yield anything recognizable as matter.
So we've got parameters that physics itself establishes with respect to the range of the constants.
And within that finite range, then we've got ranges that are life permitting and they are in fact calculable and in many cases turn out to be very small.
Right. So you perturb the parameters of the standard model and general relativity on paper and you can watch the universe fall apart. And physicists have done this.
So the slice in which life is possible over the established meaningful physical range that is provided roughly 40 orders of magnitude. If you take the gravitational force strength through the strength of the strong force as one way of quantifying this, it gives you 40 orders of magnitude of strength.
The slice of that in which you got a life permitting condition coming about as a result of the value of the constant is very small and the constants fall in that range. So that's one way you can address the question. You don't need an infinite sided dice. You got a finite range that's definable by physics itself.
[00:19:49] Speaker C: Thank you. Now, another pushback was this idea of Hitchcock's bean jar idea. Could you explain what that means and why it's not a good response to the fine tuning argument?
[00:20:02] Speaker A: Yeah, sure. So Christopher Hitchcock is a philosopher of science at Caltech and he proposed an explanation of various things in nature. And Holper suggests it might apply to fine tuning. And so what is the example of the bean jar? Well, suppose you put like lentils and black beans into a jar and shake it. Okay, Initially they're all mixed up, but they sort into layers and the arrangement looks like it's wildly improbable, as if it had happened by chance.
But gravity and difference in densities does the work to sort it as you're shaking the jar.
And so the moral that's drawn from that is that an orderly and improbable looking arrangement doesn't require a desire.
Perhaps fine tunings like that, you know, shake the jar of the universe and things fall out.
There's a natural sorting mechanism that we haven't yet discovered that does this for us. Well, so let's try to take that seriously as a suggestion.
And if we do, I think it actually helps our side of the debate. More than it helps Phil's side of the debate. So what, what's, what's happening?
We're not surprised by the fact that improbable things happen.
And the identification of a causally adequate mechanism, in this case gravity and density sorting, is what makes this outcome expected, even though it might initially have seemed surprising. All right, so the improbability of the situation was answered by a cause. We provided a cause that explained what was happening. And that, that seems to be the lesson of the bean jar. It's a lesson about explanation and about how we go about looking for an explanation. Now, if we apply it to the fine tuning situation, okay, fine tuning has that structure. It's improbable, it's independently specifiable as an outcome, and we're looking for a cause. All right, now here's the crucial difference between the bean jar and this situation.
The fundamental constants are kind of at the bottom of the explanatory pile. They're at the bottom of the physical stack. There's no deeper physical mechanism underneath them that's doing the sorting.
That's what it means to call them fundamental constants. Right? So the bean jar doesn't retire the question.
It's not as though there's a jar that's being shaken here that the, the constants are going to fall out of. They're, they're already at the bottom of the jar, all right, so to speak.
So what it sends us looking for is a cause. All right, and so what are the candidate causes for explaining these things? Well, one could be physical necessity.
The constants couldn't have been otherwise than the way they are. All right, well, that's coherent, but current physics doesn't give it any support whatsoever. There's almost nobody in that camp. There are a lot of people wishfully thinking, wouldn't it be wonderful if this were the case? But it's not the case.
So what else can we do? Well, we can say there's a multiverse. It doesn't remove the tuning, but it relocates it to a universe generating machinery. All right, so you got a bunch of different universes with different initial conditions and different constants and so on, so forth. And you have an observer selection effect. That explains why we see what we do. But the problem is you've really shifted the fine tuning into the universe creating mechanism itself.
It's got to be constructed in a certain way if it's going to produce universes. And so it's got to operate in a lawful way with law like constraints, as well, you're not getting outside of explaining where the laws come from or where the constants come from, because you've pushed them back into the multiverse generating mechanism itself, whether that's string theory or M theory or whatever you're going to use to try to generate a universe with different laws and constants.
Okay, so what else could we do? We could say, well, it's just brute fact, right?
Doesn't need an explanation.
So you're really offering an explanation. You're abandoning the explanatory demand altogether. And that sits badly with the principle of sufficient reason.
That would say that every contingent state of affairs has an explanation for being the way that it is and not some way else.
And if you deny that, there's a path that leads to really an overwhelming skepticism. And it also destroys the foundations of science, because science is about the business of offering explanations for things. And if you stop doing that, then, I mean, I guess you could be an instrumentalist, but there are difficulties with that position as well.
So what's left? We've physical necessity won't do it, Multiverse doesn't do it. Brute fact is not an option.
So we're left with a mind as a possibility.
It's one candidate that's independently known to be causally adequate to the production of finely tuned specified systems. And we watch minds do that every day, right, Than the human sphere.
So when we see it in nature, it's got to give us pause, especially when there don't seem to be any other viable alternatives that are causally adequate to the explanation. So anyway, Hitchcock's bean jar leads you to a cause. And that cause, if you're dealing with the fine tuning of the constants of nature, points you toward mind, ultimately, I would say.
[00:26:16] Speaker C: Well, that's a wonderful response. Now, we've talked about why the fine tuning argument is valid.
Can you explain how do you actually determine the level of fine tuning? What's the rigorous approach to doing that? And could you give us an example of that calculation?
[00:26:33] Speaker A: Physics gives us measurements, and in respect of fine tuning, all right.
And each of these measurements, you could say, does two logically distinct things. All right? So a single measurement tells us how small the life permitting window is for a given constant. And this does two different jobs, and keeping them apart helps to make the subject clearer. The first job is one of comparison.
So we can ask how much better does design explain the evidence than blind chance?
All right?
And there's a way of dealing with that in Bayesian analysis that involves something called the Bayes factor. And I'll say more about that in a minute here.
And the other approach, the second job is a kind of chance elimination. Can chance be ruled out on its own terms without mentioning design at all? Okay.
And the way of doing that is a concept called specify complexity. Okay, so first of all, let's talk about the comparison job. Let's talk about the Bayes factor. So Bayes theorem, if we put it in odds form, it's kind of a ratio of two evaluations.
Posterior odds of, say design over the posterior odds of undirected chances equal to the prior odds ratio times the Bayes factor. And that factor is a ratio of likelihood likelihoods, the probability of the evidence given design divided by its probability given chance.
All right, so if you're looking at just the Bayes factor, it's the ratio of the likelihoods and we can set aside the priors. We don't need to worry about our intuitions ahead of time as to what's the best explanation and why, which direction we're inclined to toward. We're just going to look at the Bayes factor.
It's priors free by construction.
And this will eventually take us back to your remark that it comes down to the priors in the sense that once you've got the Bayes factor in place and you've got a quantitative analysis there, you can tell how extreme your priors have to be in order to get a certain conclusion, if you want that conclusion as your posterior. All right, so the numerator and the denominator of the Bayes factor behave differently on design. You've got a life permitting universe and it's unsurprising.
So the numerator in terms of the likelihood only has to be non negligible. All right, but on chance, a life permitting universe, given what we know about the fine tuning is incredibly unlikely.
And so the denominator is minuscule and the ratio then is enormous. You get a very large Bayes factor. In the meantime, the number in the denominator we get from not Bayesian statistics, but Fisherian statistics.
So we specify blind nature, assign it a non informative distribution and ask whether the observed result lands in a vanishingly small sliver of possibilities.
All right, so that sliver's size is exactly what the Bayesian comparison needs. So Fisher supplies, if you like, the raw data and then Bayes uses it. These aren't competing statistical methods in this context. They're allies, they're not rivals.
All right, so that's the idea of the Bayes factor. Now, I described the first job. The second job I said was chance elimination. So let's talk about that specified complexity.
So we've got a structural fact, we've got a Fisherian test, that's a one hypothesis test, it's not a comparative test and it asks only whether chance is adequate. Okay, doesn't ask what design would predict or anything like that. It says you've got the same number that can stand alone and convict chance without reference to the numerator at all. All right. It's a verdict available even to somebody who refuses to entertain design.
It's the chance hypothesis counting against itself, so to speak. So what is specified complexity? Well it's if you like the difference between complexity, which is the improbability of the target in bits, say base 2 and specificity or specification, the cost of naming that target in advance.
All right, so if what's left is large after you subtracted the specification cost from the complexity of the situation analyzed in the first place, if what's left is large, then chance can be decisively rejected and the guarantee is distribution free.
You might note at this point it only needs a defensible probability in respect of the complexity observed plus a way of talking about the specification.
So it's not the global shape of the distribution of a possible universes that it it at all. You don't need to worry about that. The, the question of inscrutable dice just goes away.
All right, so one thing to keep in mind with the Bayes flat factor in the denominator, what you've got is the complexity calculation, not specified complexity. Once you've subtracted specification, you've exited the door of a comparative explanation and, and you're taking a standalone test against chance. So in terms of an evaluation then of specified complexity, we've got a Google to one basically is a ratio explaining how much specified complexity, how ill prepared or inadvisable it is to think that this happened by chance. So it's chance itself that has spoken against itself.
Design plays no part in the calculation. Chance is convicting itself here, so to speak.
[00:33:18] Speaker B: So Brian, what would you say to this? He said a Google to one that caught my attention just now.
And Bruce, again it's amazing that you can break this down in such detail and I really enjoy giving our listeners and viewers just some of that technical detail. Brian, could you just bring it home for us here as we wrap up today?
What is the significance of what Bruce is mentioning in terms of probabilities here?
[00:33:48] Speaker C: Well this is very significant because what you find is none of the Arguments against the evidence of design in the laws of physics are valid. Each of them has serious logical problems.
And also what Bruce has articulated very, very rigorously is that the probability of the laws of nature having the right values to support life is infinitesimally small. It's just unimaginably small.
So the only real explanation left is that there is a mind behind our universe that designed it for the purpose of supporting life.
[00:34:23] Speaker B: Right. And we could say that, that that isn't the 100% proof. We can all go home, but it's at least a stronger explanation than what else is on the menu, so to speak.
[00:34:35] Speaker C: Well, I think that the beautiful thing about the Bayesian analysis that Bruce has talked about is that what it shows you is that if you're even remotely open to the possibility of a designer, then the evidence demonstrates very, very strongly that there is a designer. And the reason certain physicists like Phil Helper and science communicators deny the evidence of design is not that the evidence of design isn't clear, but it's because they assume from the beginning there can't be a designer. So if you assume there can't be designer, it doesn't matter how strong the evidence is for design, they're going to reject it.
[00:35:13] Speaker B: And that's a very important point and a good one to end on. Their commitment to methodological naturalism prevents them from even allowing mind in the door. Therefore, they got to turn to what else is on the menu and endlessly deliberate over which one might work. But they're forgetting the elephant in the room.
Or perhaps they've shooed the elephant away, but either way, it's still there. Well, I think that's plenty to chew on for our first segment together, but we're going to come back in the second episode and actually turn our attention to the method that Bruce, you've developed for evaluating these cosmological models. It's a scorecard, and we're going to look at how it works, the criteria it uses for evaluation, and then we'll run, run through it with a few of the different models just so people can get a good idea of how it works. So, Bruce, Brian, thanks for your time today.
[00:36:05] Speaker C: It's been a pleasure.
[00:36:06] Speaker A: Thank you, Andrew.
[00:36:08] Speaker B: Now, audience, if you're interested in learning more about this topic, a good place to start would be the debate with Phil Halper and Stephen Meyer. You can find that on YouTube. And of course, Stephen Meyer's book, Return of the God Hypothesis is also a great tool. He does dive into this to some degree. And that's a good place to start as well. So we'll mention a few other resources in part two of this conversation. But for right now, check out the debate, substantive and civil and perhaps worth your time for sure. And then, of course, Stephen Meyer's return of the God Hypothesis. And we'll be back to conclude this intriguing conversation. So stay tuned for I Do the Future. I'm Andrew McDermott. Thanks for joining us.
