Last month I caught up on an old thread with On the Bayesian Reversal of the Fine Tuning Argument by Sober, Ikeda, & Jefferys (against Barnes & Lowder). Luke Barnes has now thrown up a bunch of responses that are even more bizarre. One of the things I observed is how he never addresses any of my actual arguments. And now he keeps doing this yet again. And I think he sincerely doesn’t even know this is what he is doing. It looks like he delusionally believes I argued things that I didn’t, and delusionally doesn’t see the things I did argue, even when I explain them to him. I don’t know how to interact with someone like that. And on top of that, now he seems to be contradicting himself and isn’t aware he is. This is genuinely strange.
Because continuing this looks impossible—Barnes has so consistently ignored what I actually say, that I do not see the likelihood of his ever actually responding to me, making any further engagement a waste of my time—this might be the last time I bother addressing him. I’m giving him one more shot only because he’s supposed to be an actual cosmologist and not some rando. But be aware, yet again, he is already refuted by everything I already actually wrote in the original TEC article and in my latest reply to him (with one exception I’ll get to below). So honestly, you could just go back and read those. That’s all you need to see how irrelevant or wrong everything he keeps saying is. But I’ll survey it anyway.
I shall be ignoring everything he says that he doesn’t back with any argument or evidence (all his complaining, for example; all his unsupported assertions, about me or what I claim or argue; and so on). He makes a lot of assertions (such as that I didn’t do one thing or another; or that he doesn’t have to address one argument or another) but never shows this is the case. Those claims I am disregarding as undefended. You should too. If he doesn’t back a claim, if he just asserts something, you should discard it as unverified. But some of his claims he at least sort of tries to justify. Those I’ll address point by point.
Parts One & Two
Barnes has replied with a post in several parts. Here are the only points he makes in entry one that he presents any case for:
(1) The first argument Barnes makes is:
Deriving frequencies from reference classes is trivial—you just count members and divide. The problem that references classes create for finite frequentism is their definition, not how one counts their members. So, Carrier doesn’t understand the reference class problem.
This is not true. I discuss the problem extensively in Proving History (e.g. pp. 229-56; see also index, “reference class”). Once again, Barnes ignores everything I actually say, claims I said nothing about it, and makes a false assertion in result. His link is irrelevant. He never explains how this problem affects my argument in TEC. He also doesn’t explain how it isn’t the exact same problem for proponents of the fine tuning argument. Consequently, his first argument lacks any substantive content relative to this debate.
(2) The second argument Barnes makes is:
Carrier’s approach to probability is inconsistent. He keeps shifting the goalposts. In TEC, when talking about a cosmic designer, he says “Probability measures frequency (whether of things happening or of things being true)”. Only known cases, verified by science, can be allowed in a reference class. But now, in OBR, it’s OK to put hypothetical possibilities in a reference class.
This baffles me. Because if he actually thinks you can’t put hypothetical possibilities in a reference class, then he must conclude the fine tuning argument invalid. Because if you can’t put hypothetical possibilities in a reference class, you can’t put hypothetical universes in a reference class. And if you can’t put hypothetical universes in a reference class, you can’t make any claim about the frequency of those universes that would or would not bear life. Ironically, he almost is here getting precisely the point I myself made when I cited the McGrews paper at him demonstrating pretty much this very point: that the fine tuning argument is a non-starter because you can’t generate any usable probability for it.
Notice, once again, this is yet another argument of mine that Barnes ignores.
Notice, also, that my argument in TEC does not require solving this problem. At no point does it involve calculating the frequency of universes that would bear life among all possible universes. I don’t even do that in my multiverse argument (see below). That Barnes doesn’t notice this, is just about the Platonic ideal of what I’m talking about: Barnes is not even addressing anything I have ever argued!
(3) The third argument Barnes makes is an extension of that one:
This destroys his argument on page 282-3 of TEC, in which Carrier distinguishes cases that science has verified from “alleged cases”, which must be excluded from the reference class. But alleged cases are logically possible, so they should have been included all along, according to OBR.
This makes no logical sense. Once again, he thinks somehow I am running frequency estimates in the later argument (about fine tuning), when I am not (in fact, that we don’t have to is the argument). But worse, he ignores the actual conclusion of the section he is now talking about, which is that the prior probability the universe was intelligently created I will regard for the rest of the chapter to be 25%! Barnes is not even arguing here against my conclusion! He is therefore not arguing against anything I actually said. Again.
At my most charitable, I can suppose maybe Barnes means this as a meta-argument, not against anything I actually argued (about what prior to use in the fine-tuning argument, or later about the fine-tuning argument), but against what he thinks is an inconsistency in how I estimate frequencies in general, even though it has no bearing on anything we are debating. But even then his statement is bizarre. When I talk about the prior on pp. 282-83 of TEC, I mention we have to exclude “alleged cases” of confirmations of supernatural agency from our knowledge-base, because we don’t know they were or would have been confirmed, so we can’t count them as having been confirmed—a fundamental axiom of science. Why Barnes is now against science, I don’t know. But Barnes is actually now saying that because it is logically possible that we might be able to confirm instances of supernatural agency someday, that therefore when we ask how likely that is based on past cases, we should count all supernatural agency claims as already having been confirmed to be actual supernatural agency claims! That. Makes. No. Fucking. Sense.
I cannot fathom why Barnes would ever say this.
Probabilities must be based on knowledge. Even in hypothetical sets, things we don’t know about the set have to be excluded as not being known.
Notably, this is what I keep explaining to him about all the physics papers on fine-tuning he cites: they lack sufficient information to make usable claims for a fine tuning argument. Barnes never responds to any of my arguments on that point. He just gainsays the point, as if I made no arguments.
(4) The fourth argument Barnes makes is:
A fine-tuned universe is 100% expected on atheism if and only if observers are 100% expected on atheism. Observers are not 100% expected on atheism, because most possible universes do not support observers—that’s the point of fine-tuning. Thus, a fine-tuned universe is not 100% expected on atheism.
Technically this I should include in the set of mere assertions. Since there is no actual argument here. Just an assertion. But I include it as a representative example of an assertion I refuted in detail in TEC and again in my latest reply. Barnes just keeps ignoring everything I said and claims to have refuted me. He can’t have, since he never even addresses any of my arguments. Which already refute him. This is maddening. Why do I bother?
One more time:
Yes, less than 100% of possible universes will contain observers; but those other universes will never be observed; therefore, we can’t be in one. The fact that we are observing, entails we are in an observable universe. So we cannot use that fact to distinguish what caused this universe to exist. All atheist universes with observers in them will be fine-tuned. Thus, since fine-tuning exists in all atheist universes, it cannot be evidence against atheism. We thus are left to decide on the prior probability. Which is an entirely different question. One that does relate, for example, to what Barnes dismisses as the threshold probability of fine tuning, but he shows no signs of understanding what that’s based on or why it’s relevant. Nor does he propose any alternative to it. He also shows no signs of understanding the relevance of prior observations as well, such as I discuss in TEC, pp. 282-83. Again, Barnes is simply dismissing and thus ignoring everything I argue.
I’ll also reiterate a point I made last time, that still seems to elude Barnes, that “less than 100% of possible universes will contain observers” is actually just as true on theism as on atheism. Since the assumption that God would ever or only make inhabited universes is actually unjustified by any evidence or logical entailment. So you have to arbitrarily choose a God with the requisite motives out of the set of all possible gods to get observers on theism, just as physics had to arbitrarily choose a universe with the requisite features to get observers on atheism. And so far as we know, it’s a wash. We can’t show one more likely than the other (as I explained in detail in my last reply). Worse, it just so happens, that more God universes aren’t fine tuned than atheist universes, because gods don’t need finely tuned universes. Another argument Barnes keeps repeatedly ignoring.
I know it’s futile. Barnes ignores everything I say. But for my non-delusional readers, this is how it breaks down. When ~e = a non-fine-tuned universe, then P(~e|God) is more than 0. Just for illustration, let’s say it’s 60%, since there are so many better universes a God could make without the limitation of finely tuning obscure physical constants. So imagine for a moment that 60% of all the universes God could make do not require such fine-tuning. Thus, P(~e|God) = 0.60. But if P(~e|God) = 0.60, then it is necessarily the case that P(e|God) = 0.40, since P(e|God) is necessarily the converse of P(~e|God). And P(e|God) means P(fine tuning is observed|God). So on this illustration, fine tuning is only 40% expected on theism. It is by contrast 100% expected on atheism. Because P(~e|Atheism) equals flat out zero. If atheism is true, then there are no universes we could ever observe ourselves in that would not be finely tuned.
All the objections Barnes tried to raise against this were already in fact dealt with in TEC, as I explained yet again in my last reply, where I reiterated and explained those objections. Barnes still ignores every single one. He thinks he isn’t ignoring them. He keeps making arguments. But none of them address any of my arguments. Barnes has still not even indicated that he understands what my arguments are. Because he thinks he knows what they are. But he is arguing with a delusion of me in his head. Not with the actual arguments I have made. And there is no point arguing with someone like that.
(5) The fifth argument Barnes makes is:
Carrier’s discussion of the multiverse uses a different approach to probability, one that is inconsistent with the approach to probability applied to fine-tuning elsewhere in TEC. This inconsistency undermines his entire approach—the goalposts shift at will.
Ironically, now Barnes is moving the goal posts by changing what his argument was. Originally it seemed he was trying to make an argument about our supposedly being unable to develop a probability distribution over an infinite space for the purpose of calculating the probability of a life bearing universe among all possible universes—a point that in fact I have agreed with (I even cited a formal paper demonstrating it)—but anyone who concedes it, must concede the fine tuning argument is invalid. Barnes does not seem to recognize this consequence. Even though I have said it to him multiple times now. And yet he contradicts himself elsewhere by insisting we can develop those infinite probability distributions. He can’t decide which it is.
Again, nowhere in TEC do I use an infinite probability distribution. It seems like Barnes is admitting this now. So maybe he will confess that this argument of his has no relevance whatever to my argument in TEC. To again try to be as charitable as I can, I will assume Barnes meant to be criticizing a completely different argument of mine (which I completed here). There, I do not attempt to calculate a probability! Rather, I show that whatever probability would result given a certain set of assumptions (which assumptions I clearly state), it must be infinitesimally close to 100% that there would be a life-bearing multiverse.
To head off Barnes-style misreading here, let me reiterate that that argument is only based on stated assumptions, and explicitly says the conclusion does not follow if we abandon one of those assumptions; it also shows the consequences of abandoning any of those assumptions. And also, one must be clear, that to say we don’t know how to calculate that probability is not the same thing as saying that we can say nothing at all about that probability. This is easily shown by simply doing it all with finite probability distributions, and then using the method of exhaustion (increasing the finite number of universes logically possible) to show near where the probability is going to be, even though we can’t get all the way to precisely what it is. Barnes never addresses any of these points.
So, not only does this have nothing to do with my Bayesian argument in TEC, and not only it does it not use Bayesian arguments at all (and so is irrelevant to any debate over my use of Bayes’ Theorem), it’s a completely different argument based on completely different premises, not a single one of which Barnes interacts with.
Now, having moved the goal posts himself, Barnes accuses me of moving the goal posts with his bizarre claim about me being inconsistent somehow. This wasn’t his argument before. But stymied on that, he has switched gears and is making it his argument now. That accusation is false. Which is bizarre enough. But it’s even more bizarre because it makes no sense to say I can’t use different interpretations of probability in different arguments. As long as I am consistent within each argument, there is no inconsistency to complain about.
I don’t use different interpretations. But even if I did, that’s perfectly legitimate. Unless Barnes means to insist that only one interpretation is valid and anyone (not just me) who uses a different one is always wrong. But weirdly, Barnes seems to be trying to insist upon the opposite, that I am wrong to insist on only one interpretation (that it all reduces to frequency). So Barnes is contradicting himself. On the one hand, he doesn’t like that I argue that all interpretations of probability reduce semantically to some statement or other about a frequency. And on the other hand, he insists I not use more than one interpretation of probability, because doing so is somehow “being inconsistent.” Baffling. Truly utterly baffling.
But it’s also just plain wrong. In my multiverse argument, I do not use anything but a frequency interpretation of probability. Why he thinks I’m not, totally escapes me. On the one hand, he seems to think hypothetical universes can’t be counted (which refutes all fine tuning arguments—thank you, Luke Barnes!). On the other hand, he seems to think they can be counted and that that count will be small for life-bearing universes against all non-life-bearing universes. Which, as I explained to him, the McGrews refuted. But remember, Barnes never responds to any arguments I make or cite.
Barnes is the one contradicting himself here. Not me. Pick a lane, Dr. Barnes. Can hypothetical universes be counted to generate a frequency of life-bearing universes among them? Or not?
(6) The sixth argument Barnes makes is:
There is no such thing as “transfinite frequentism”. Take a moment to Google that phrase – the only result is Carrier’s blog post (and possibly now this one). Literally no one ever – no mathematician, no scientist, no philosopher … not even a clueless quack – has ever used that phrase before, so far as Google (and Google Scholar, Google Books, Wikipedia, arxiv.org, Bing, Yahoo!, and even Ask Jeeves) can tell. Draw your own conclusion. The two kinds of frequentism are called “finite frequentism” and “hypothetical frequentism”.
This is really a face palm moment. First, let me google that for you. And then google “transfinite probability” since remember, I am talking about Barnes’s claim that transfinite probabilities cannot derive from transfinite frequencies. He’s the one claiming you can’t do “frequentism” on infinite sets. I then responded by saying, neither am I, so what’s your point? Barnes then ignores what we are actually talking about, flips his lid over a mere coin of phrase I used to communicate that to a popular audience, and then thinks “hypothetical frequentism” isn’t what I was talking about, when obviously it was. It’s just that readers won’t know what that means or what connection it has to transfinite mathematics, the point Barnes was trying to make (a point that’s going over most readers’ heads), so it wouldn’t have been a useful term to use. Scientists tend to have a real hard time understanding the importance of mass communication and that most readers are not going to know what their jargon means. And I write for a mass audience. Get over it, Luke Barnes.
So let’s stop flipping our lids over pointless bullshit, and let’s talk about what we are actually arguing. Barnes is claiming we can’t get probabilities by counting frequencies on infinite sets. In what sense that is or isn’t true is a whole other debate. But it doesn’t matter. Because I don’t do that in TEC. So it’s moot. And I don’t even do it in my multiverse argument. What I do there follows by method of exhaustion, just like all other hypothetical frequentism: I show that as the number of possible universes approaches infinity, the probability of getting a multiverse with at least one life-bearing universe within it approaches 100%. That is not stating what the probability is. It’s stating a general area in the overall probability space it must lie. Barnes has never even described this as my argument, much less rebutted it. Once again, Barnes would rather whine about terminology than ever even respond to anything I’ve ever actually argued.
Barnes also references cool articles by Alan Hajek against finite and hypothetical frequentism. And once again, Barnes does not interact at all with my defense of hypothetical frequentism (in Proving History, pp. 257-65). He doesn’t say how the cited article is even a rebuttal to that. For instance, my discussion of physical-hypothetical modeling (a generalization of exactly what Barnes says physicists do to get frequencies of fine tuning) refutes a third of Hajek’s points, and substituting the method of exhaustion refutes another third of Hajek’s points, and the remaining third doesn’t relate to anything we are discussing, because they only pertain to the difficulties specifying infinitesimal probabilities (a problem Hajek himself even proposes to solve), and I have never relied on specifying an infinitesimal probability. This is just another example of how Barnes ignores everything I say. And makes zero effort to understand it.
Part Three
Here are the only points Barnes makes in entry two that he presents any case for:
Actually, apart from one exception, there are no points he makes in entry two that he presents any case for. He just makes a bunch of assertions, none of them supported. And ignores much of what I said as well.
So all I was left with to respond to (for the rest, what I have already actually argued is sufficient response) is actually the one example I’ve yet encountered where Barnes actually responds to something I said, and correctly rebuts it!
Barnes shows (through linked work) that Fred Adams’ finding of a 25% habitability space within variances of physical constants should be dismissed from this debate, because Adams only meant within an arbitrary space of all universes that could have stars in any conventional sense, not all possible spaces (much of which, Adams concludes, will have stars in some unconventional sense, but we can’t be sure such universes would produce organized life). I am now convinced Barnes is correct on that point (and have added a note to my last reply accordingly). I’ll assume Barnes has similarly good arguments against the others who come up with similarly large spaces of habitability (though I didn’t check them all).
This then gets us back to where we started:
My argument in the article was, “We actually do not know that there is only a narrow life-permitting range of possible configurations of the universe.” Barnes can cite no paper refuting that statement. I give two reasons why. Barnes pretends I only gave one. And then when he gets to the second, he forgets the relevance of my second argument to the first. Only one of my two arguments for that general thesis (that we don’t know) is that some studies get a wide range not a narrow one. … Then I go on to give the second reason, which is that even those papers are useless. Because my very next sentence, the sentence Barnes hides until later, and pretends isn’t a continuation of the same argument, says: “And even those models are artificially limiting the constants that vary to the constants in our universe, when in fact there can be any number of other constants and variables, which renders it completely impossible for any mortal to calculate the probability of a life-bearing universe from any randomly produced universe. As any honest cosmologist will tell you.”
The first argument I now concede Barnes has a case against. But that leaves the second. That second argument consists of two points: one is the mathematical impossibility and the other is the parameter problem.
The math problem Barnes simply insists he can get around—but all he does is handwave; he still never addresses or rebuts the McGrews arguments on this. Their arguments thus stand unrebutted. Instead Barnes uses the most astonishing ad hominem fallacy ever: he quotes the McGews trash-talking me on a blog, as if that had anything to do with their being right about their peer reviewed published paper refuting Barnes. Really. Now, what the McGrews say about me on their blog is hardly an honest and credible treatment (you can see that for yourself—it’s also six years out of date). But it’s still completely irrelevant to what Barnes and I are debating. Barnes can’t avoid my arguments by claiming someone else thinks I suck. And for him to use that as an excuse to avoid the arguments of even them, the very people whose judgment he thinks we should trust, is some chutzpah. It seems he is just intent on avoiding their paper and its implications to his case altogether. And hopes you don’t notice that’s what he’s doing.
Then there is the parameter problem. The number of possible parameters is not limited to the parameters in our universe. In our universe, many of those parameters may be zeroed out or so small as to be undetectable, or not even physically possible yet possible in universes of other configurations. And since I wrote this article, numerous leading cosmological physicists went on record siding with me on this, so Barnes is pretty well cooked here. I’m voicing the expert consensus. He’s ignoring it.
For example, Adams only ran the math for the parameters that exist in our universe. But this does not address universes that have other parameters. Adams explored what happens when you vary with respect to each other four forces: gravity, electromagnetism, and the strong and weak forces (which two he combines into a single factor). But what if in our neighboring universe, let’s say, there is a fifth force that changes how these forces interact? Indeed what if that fifth source is repulsive instead of attractive? What if it is only carried by some particles and not others? What happens then? Will some of those universes contain stars and thus chemistry and thus life? How are we supposed to claim to know? We can’t. Similarly, our universe has three open dimensions of space. What about universes that have four or ten or a million? What happens then? Show me the paper that covers this, and the infinite number of possible other forces, and everything else.
Then what if when the gravitational force increases beyond a certain point greater than the other forces (however many forces there happen to be in this hypothetical alien universe we are discussing) that it collapses the resulting universe, causing another Big Bang? (As it surely would do, barring the introduction of yet more infinitely possible changes to the physics.) And what if that always resets the value of gravity, as it presumably must if the strength of gravity could be chosen at random in the Big Bang in the first place? Then there will be an upper limit on the gravitational constant’s size in relation to the other forces—because all universes with larger gravitational forces will keep collapsing until a smaller force is generated. So, what is that upper limit? Are there similar upper limits on the other forces? Again, which other forces? (Since there can be more than the ones we have.)
BTW, Barnes now insists “Bayesian probability deals with free parameters with infinite ranges in physics all the time.” Contradicting his claim that I can’t do probability with free parameters with infinite ranges. Sigh. But that’s still only when you know the constraints. So he is still not getting the point I am making now, and have been making since my last reply, which is that a truly infinite parameter space includes infinite possible numbers of spatial dimensions, infinite possible numbers of forces, repulsive as well as attractive, infinite possible laws of physics governing the interplay of those forces, and an infinite possible number of ways these forces could be constrained by processes that reset them. Truly, Dr. Barnes, has any peer reviewed paper done the math taking all these actual possibilities into account? I’m pretty sure the answer is no. As I said from the beginning. And Barnes simply won’t admit this.
If a universe is picked out of all randomly possible universes, what is the probability it will generate any kind of intelligent life somewhere inside it?
We don’t know the answer to that question.
Part Four
There is only one point Barnes makes in entry three that he presents any case for:
In response to my saying:
Barnes wants to get a different result by insisting the prior probability of observers is low—which means, because prior probabilities are always relative probabilities, that that probability is low without God, i.e. that it is on prior considerations far more likely that observers would exist if God exists than if He doesn’t.
Barnes says this:
Those sentences fail Bayesian Probability 101. Prior probabilities are probabilities of hypotheses. Always.
And he rightly goes on about that. But my wording misled him. So I can’t blame him here.
What he means is indeed what I meant: hypotheses of a coincidental God creating observers or a coincidental physics doing so. I should have said “prior probability of Godless observers being low.” In other words, he wants the prior probability of an observer-creating physics to be a lot lower than the prior probability of an observer-creating God. But as these are relative probabilities, the prior probability of “an observer-creating God” is not simply the converse of the probability of fine tuning, any more than the prior probability of a rich person having gotten rich by winning the lottery is the probability of winning the lottery. If there are only two ways to get rich, A and B, and each one has a probability of 1 in 1,000,000, then the prior probability of a rich person having gotten rich by A is not 1 in 1,000,000, it’s 50%. I was trying to make the same point about the hypotheses of how observers came to exist. And Barnes still has never responded to this point. Although perhaps in this case because I confused him.
A posterior probability generated by one datum becomes a prior probability for the next run of the equation. So that if we run the math for only one datum (there are observers), we can start with the priors for the two hypotheses “God” or “Godless Physics,” and generate the posterior probability of observers. And then that posterior probability is used as the prior probability when we run the math for the next datum (fine tuning is observed) for the same two hypotheses. This is what I was referring to. I was assuming he wanted to break the argument into two stages, one run on the bare fact of observers (before we consider fine tuning), and then plug the result into the run for “fine tuning.” I just worded this badly.
A wholly better way to say it is:
Barnes wants the probability of Godless observers to be low, in order to argue (perhaps?) that the probability we are God-created observers is high. And he wants to do this by saying “observers” are unlikely on Godless Physics but likely on God. And he wants to use fine tuning to do that. But fine tuning doesn’t help with that. Because fine tuning is always true when there are observers and no God. So we need something unusual about observers alone. We can presuppose a God that will likely create observers, but what’s the probability of that God among all other possible gods? Is it higher than the probability of a life-bearing universe among all other possible universes? How would we know? If you just presuppose a life-interested God, then you can just presuppose equal luck for the Godless Physics. And that gets us nowhere.
Everything Barnes thus says at this point is correct. It just doesn’t address what I was actually talking about (or anything I have discussed at all). Though in this case, his confusion is entirely my fault. Whether for that reason or not, he still has never addressed the point.
Conclusion
That’s it. That’s all Barnes had in reply to my last entry in this debate. He left most of it without response. And got wrong everything else (in one case because I misled him), except for one good rebuttal (of my citation of Fred Adams), which I now acknowledge.
Barnes summarized his points at the end. Let’s look at that recap:
- “Carrier has not addressed the charge of inconsistency with probability theory.”
Barnes has not identified any inconsistency relevant to my article in TEC or my (very and completely different) multiverse argument.
- “He has made up probability concepts that no one has ever heard of before, including ‘transfinite frequentism’ and ‘existential probability calculus’.”
Colloquialisms are not relevant to this debate. Barnes needs to address the arguments signified by the words. Not complain about how much he doesn’t like colloquial words (especially jokes, like the latter was).
- “He has abandoned his previous claim that ‘all the scientific models we have … show life-bearing universes to be a common result of random universe variation, not a rare one’.”
This I now agree with. I will correct the article in which I said that.
- He completely misunderstands my rather obvious point that ‘for a given possible universe, we specify the physics’, and in so doing, shows that he does not understand fine-tuning at its most basic level.”
Actually it is Barnes who doesn’t understand what I am talking about. You don’t get to just decide which universes are physically possible. There are more possibilities in the possibility space our universe would be randomly chosen from than universes with only and all the same parameters as ours. Universes can have different shapes, dimensions, forces, physics, they can have repulsive forces as well as attractive, they can have physical limits on how strong the forces in them can be, and many other physical differences. Barnes has consistently failed to grasp the significance of this point. Any argument of the form “observer-making universes are extremely few in the set of all the universe that could randomly exist without a God” cannot be made. Because we cannot even define all the possible universes in that set. We certainly have never counted them all up and determined how many would make observers possible. I’m pretty sure none of the papers Barnes cites has ever done that. I am pretty sure it’s impossible to do. The number of possible ways to vary a universe is multiply infinite (infinitely many dimensional structures, infinitely many arrays and combinations of forces, infinitely many ways force maximums can be limited, and so on). And the McGrews paper on the mathematical fact that this can’t be done, either, stands unrebutted. By Barnes or anyone.
- “And, finally, Carrier’s argument regarding the ‘Real Heart of the Matter’ is rendered meaningless by a deep misunderstanding of probability theory’s basics.”
Barnes actually did not show that. He correctly identified some poor wording at one point. But correcting the wording doesn’t change the argument there. And he still has never rebutted the argument.
Where things stand right now:
- Barnes has never rebutted my actual arguments in TEC. He has gainsaid it. But he still won’t respond to the arguments in it that rebut him.
- Barnes does not even seem to understand my multiverse argument. He has yet to correctly describe it, or rebut any of its premises.
Barnes seems to contradict himself by denying we can use hypothetical frequency sets, then insists we use hypothetical frequency sets to calculate how many hypothetical universes would bear life; and by denying we can derive frequencies from infinite sets, then insisting we can derive frequencies from infinite sets. He needs to pick lanes here. Can we use hypothetical frequency sets to calculate how many hypothetical universes would bear life? Then we can use hypothetical frequency sets. Can we derive frequencies from infinite sets? Then I can derive frequencies from infinite sets. Although I never have. Neither my argument in TEC nor even my multiverse argument (which alone even discusses infinite sets) does that.
What Barnes doesn’t get is that fine tuning will be observed in all observed atheist universes. Therefore P(fine tuning is observed|atheism) = 1. This is not the case for God-made universes. P(fine tuning is observed|God) is actually less than 1. Because on theism (and only on theism) could we ever have observed ourselves in a non-finely tuned universe. This brings us to the question of the basic probability that there would be observers in any sense at all (apart from God). This depends on the relative probability of getting a lucky God or a lucky universe; not the absolute probabilities of either. And the probability of fine-tuning a universe is an absolute, not a relative probability; that means we still have to put it in ratio to the probability of the requisite fine tuning of a God. And we don’t know what that relative probability is. No matter how low fine-tuning may be as a probability, the luckiness of our God may be just as improbable. We literally don’t know.
And note that it can’t be just a God lucky enough (for us) that it happened to want to create any kind of observers—since we observe we aren’t angels in heaven, for instance—but who wanted to create observers of our specific, messy, physical, unheavenly kind, in an almost entirely inhospitable universe, a universe built in exactly all the ways a Godless universe would have to be. In other words, a God who not only luckily exists, and who was not only amazingly luckily possessed of all the requisite knowledge and powers and desires, but a God who had the very specific and peculiar desire to create a universe that looks exactly like a universe without a God in it.
That’s the argument of TEC. And Barnes has still to this day simply ignored it.
Barnes is a crazy person. Simple as that.
I don’t think that’s a useful assessment. He’s a frustratingly obtuse person. He might be deluded about some things. And he doesn’t cope with colloquial English comprehension well. But that may be all there is to it.
Hi Richard,
I don’t really want to intrude on this discussion because it is about Lukes reply and I think we will just get bogged down in the meaning of frequencies again, however following up on our past discussion on fine-tuning, I have tried to make a response to your three last points (in the previous thread) here:
https://www.scribd.com/doc/296697791/Richard-Carrier-s-rough-fine-tuning-argument
where I also try to lay out the alternative way I think the fine-tuning argument should be analysed and why I think the result of Ikeda and Jeffreys is not relevant when assessing the existence of God with an example, naturally I would be very interested in your response. Regarding the last two questions I asked you on the past thread, I tried to send you an email to your infidels address but perhaps I got it wrong? The main content of the email was:
To recap:
O : That we (intelligent, observing life) exists in the universe
H: Any hypothesis
b: Background information not relating to the above two variables.
(eq.1:) p(H|Ob) = p(HO|b) /p(O|b) = p(O|Hb)p(H|b)/p(O|b)
Main Question1: Do you accept that the above decomposition eq.1, i.e. expressing p(H|Ob) in terms of p(H|b), is valid? I.e. is that something we are in fact “allowed” to do when considering the probability of p(H|Ob)?
Main Question 2: Would you agree eq.1 follows from the rules of probability theory? That is, if we are not allowed to use p(H|b), this must necessarily be because of philosophical (if you will) reasons?
(Bonus) Main Question 3: Could you provide your version of the fine-tuning argument as used in TEC which leads to your particular numerical result with an equation that uses the notation I apply above? (O, b and Ft for fine-tuning)?
Cheers,
/T
Questions All: Of course! You can do all those things.
The question though is what then is that a statement about? If you are asking the probability of what would be observed if there are no observers, you have not a validity problem, but a results problem. Because P(observations|~observers) is always zero. So if you take observers out of b, you always get zero as your result, for any possible observation (since every e is an observation).
So you can try asking after P(observations|{either observers or ~observers}), although if you’ve taken observers out of b, that can’t be what you are asking, because any “either a or b” includes a and b. So you are still assuming observers exist, in order to ask how probable it is that observers exist on e = “observers are observed to exist.” And that’s always 1, i.e. P(observers are observed to exist|observers) = 1. That’s the lesson of Descartes.
Even if you do the “either” method, you get P(observations|{either observers or ~observers}) = {P(observations|observers) x P(observers)} + {P(observations|~observers) x P(~observers)}, but that’s always just P(observers), because P(observations|observers) = 1 and P(observations|~observers) = 0, so the second half disappears (because 0 times anything is 0), and the first half is just 1 x P(observers), which is always just P(observers). And yet, because the ~observers option zeroed out, per Descartes, P(observations|observers) = 1.
The lesson?
P(observers) is always 1 when anything is observed.
Listen to Descartes.
There is nothing invalid about that. But it’s not a useful conclusion if you want to prove anything about god or physics.
This is what I’ve been trying to explain to you over and over again. (Barnes meanwhile simply dropped this whole point of debate. He never addressed it.)
So what can we do?
We can talk about the relative probability of the different ways this observation (that observers exist) can have come about. The probability that observers exist is 1; we’re here, after all. So the question becomes, how did we come to be here?
This is where that lottery analogy becomes pertinent.
Suppose we observe someone is rich (and assume it’s logically impossible to be wrong about that, as it is logically impossible to be wrong about the existence of observers). So we know the probability that they are rich is 1. But how did they get rich? Suppose somehow we know there are only two possible ways, A and B, and each has an independent probability of occurring of 1 in 1,000,000, so any given person has only a 1 in 1,000,000 chance of getting rich via A or B (and thus only a slightly higher chance of getting rich at all).
When we ask “What is the probability that they got rich by method A?” the answer is not 1 in 1,000,000. It is not the probability of A occurring!
Read that sentence again.
Okay. What then is the answer to the question “What is the probability that they got rich by method A?” It’s 0.5 (i.e. 50%).
I assume you know why. But I’ll explain for anyone who doesn’t. The probability of a rich person getting rich by A and not B derives from the ratio between the odds of A and B. Since they both are equally likely, so are A and B equally likely. Hence there is a whole 50% chance a rich person got rich via A even though the odds against A are 1,000,000 to 1 against.
So it does us no good to know the odds of A unless we know the odds of B. If we only know the odds of A, we cannot answer the question “What is the probability that they got rich by method A?”
Do you agree?
If not, you haven’t understood the point.
You can only answer that question if you know the relative odds between A and B. And just knowing the odds of A is not enough information to know that.
A is godless physics. Or we should say, any godless means of bringing about observers—we just assume the only way is a lucky choice of physical universe from a set of all possible physical universes (hence “fine tuning”), and we assume that we know what every possibility is and that we know for every one whether it would produce observers, and we assume that all possibilities are equally likely, and so on, which all are actually also not really things we know, but that’s the other problem I’ve been bringing up with Barnes—for now, let’s just assume we solved all those problems and we actually have a defensible probability for A producing observers.
B is gods. And the probability of observers on B is the probability of randomly getting a god, out of the set of all logically possible gods you could have gotten, who would want both to create observers and observers just like observers would have to look in A (as opposed to angels in a magical paradise, for example), and who would have the knowledge and power requisite to do so. Since you can’t arbitrarily declare all possible gods wise and powerful, or else you will need a new term, C, that includes all the not-solely-physical-universe possibilities that could have existed if “gods” in your new definition didn’t exist but supernatural beings still did, and C will need a probability of producing observers as well, and we’ll have a three-hypothesis equation and the prior probability space has to be divided among all three. In TEC I simply bracket all these away for convenience (as I explain in its section on priors). But if you want to do what Barnes and you are doing, you have to put all this back in. So it either goes in B or C. We’ll keep it simple and put all these possibilities in B. Then A and B exhaust all logical possibilities.
So what is the probability of observers on B?
You don’t know.
It’s pretty much impossible to even imagine what it could be.
The problems I note for getting a probability on A (which I just here laid out in reply to Barnes, for example) are nothing compared to the problems that arise for defining the entire set of all possible supernatural beings that could possibly have existed, and counting out how many could and would make observers and how many wouldn’t or couldn’t.
Since B cannot be defined, even if you can define A, you can’t know what their relative probability is.
So what Barnes and you want to do is impossible.
What you want to do is prove that there are more ways to get observers like us on B than on A. But even if we could get an answer for how many ways to get observers there are on A (and again, we really can’t, but that’s a separate problem), we can’t get that answer for B. So we are in a position analogous to knowing that people get rich by A or B, and that A has odds of 1,000,000 to 1 against, but we don’t know what the odds against B are. So we don’t know if the probability of A producing observers is 50% or less or more or what it is at all.
So all we have left are observed phenomena (e.g. how often things so far have turned out to be lucky physics and how often they turned out to be supernatural agency, to determine the odds that this next thing will be one or the other). Hence that’s what I work from in TEC. Which Barnes ignores.
So the equation isn’t the problem. Your equations are fine. The problem is what happens when you actually run the math. You either get useless results, like P(observations|~observers) = 0 or P(observations|observers) = 1, or you get an unsolvable problem, like the relative prior probability of an observer-making-A over an observer-making-B. I’ve pointed out we can’t even get a result for A. But even if we could (as Barnes thinks), we still don’t have a result for B. And you can’t know the relative prior probability if you don’t have both numbers (and, BTW, know them to a reasonable certainty; otherwise, GIGO). And it doesn’t matter if you port these probabilities into likelihoods, either (e.g. if A, then x.x% chance of observers; if B, then y.y% chance of observers), you still can’t get a result (you still don’t know what y.y% is), and you still face the question of the prior probabilities of A and B! (Which then is simply the relative prior probability of naturalism vs. supernaturalism.)
This is why what you and Barnes want to do is impossible. And thus why I had to do it the way I did in TEC.
Shivam Brahmin:
It goes without saying I agree with Richard that Luke is not crazy. If you are interested, feel free to put in your own words where Luke in your view goes wrong regarding his analysis of the fine-tuning argument and I would be happy to clarify what he is saying.
Hi there,
I’m only just recently trying to come to terms with Bayes’ theroem, and I’m making pretty slow progress, but it’s awesome!!! So I have a have a couple of questions (well, many more, but two will suffice for now.) They both relate to the penultimate paragraph of point 4 under subheading “Parts one and two”.
I’m confused in the first instance by the notation used. When you say P(~e|Atheism) = 0%, do you mean that the probobility of a non-finely-tuned universe EXISTING is 0, or of it being OBSERVED? (Pardon the caps, I’m not sure how to italicise in this comments box.) It seems to me that a non-finely-tuned universe could exist on atheism, but we simply couldn’t observe it, which seems to be an important distinction. So, if a non-finely-tuned universe could exist on atheism how would we then go about calculating the relative probabilities of each scenario? Could we, in fact?
The second thing that occured to me is a possible response that is available to theists (I’m not sure if Barnes is a theist or just a crank.) I have no idea if it’s relevant and I’m struggling to make sense of it. In response to the question “Could God make a rock so large he couldn’t move it?” the typical theist response is to declare that it isn’t logically possible, and therefore God could not do it. Could a similar argument be employed here? Specifically, in response to your statement “…since there are so many better universes a God could make without the limitation of finely tuning obscure physical constants” could the theist contend that God is precluded from creating such universes due to the constraints of logic?
I’m probably way off the mark on both points, but they seem interesting to me nevertheless.
Cheers
The latter, of course.
E and ~E are always observation statements. Even when E is the absence of a thing, it’s a statement about an observed absence of a thing. You can’t plug things not observed into the equation. If you do that, you aren’t running the equation for reality, but for the fictional world you just defined (by stating something is the case that is not).
Things you can’t observe don’t go into the equation as evidence. The variables e and b constitute knowledge. Only what can be observed is knowledge (plus what is logically necessarily entailed).
For example, if you want to ask what is the probability of my being still alive given that ravenous vampires live in my house, “that ravenous vampires live in my house” has to be a hypothesis. It can’t be in e or b. Because “that ravenous vampires live in my house” is not observed evidence. Nor is it observed background knowledge.
So if you want to ask what would be true in other universes with ravenous vampires in them, you can do that. But your results won’t tell you anything about this universe.
Thus, we can run all kinds of equations for what things would be like in universes we could never be in. But none of those results would pertain to this universe, which unlike them, is a universe we can be in.
That’s why other universes end up being unimportant. Not wholly. But in what respect takes some explaining. My last best attempt in response to Hendrix a few minutes ago might help you with that.
Only if they can produce a sound and valid syllogism proving those other universes logically impossible.
I would love to see them do that. Because that would mean heaven is logically impossible. And I can’t wait to see their head explode when they realize what they just proved.
(I do in fact suspect there could be such proofs, BTW, though they may be beyond our ken and thus never discovered; but they would refute the existence of all supernatural beings, so not really a tactical option for the theist.)
Hello Carrier,
I’ve been defending your point on my blog; I was wondering if you could say if I’m characterizing your position correctly? I summarized it as (with FT=fine-tuning, G=god, N=naturalism, L=life):
“My understanding (I could be wrong…) is that Carrier is arguing as follows:
Let us treat the denominator and numerator in (1) P(G|L FT)/ P(N |L FT) separately, for ease of notation. Then for the numerator,
P(G|L,FT)~P(G)P(L|G)P(FT|L,G)<P(G)P(L|G)=P(G|L)
where I used P(FT|L,G)<1 to get the inequality. Whereas for the denominator,
P(N|L,FT)~P(N)P(L|N)P(FT|L,N)=P(N)P(L|N)=P(N|L)
Thus,
Odds(G to N|L,FT) < Odds(G to N|L)"
Is this a fair representation of your argument from fine-tuning against god?
The entire post (which is a bit less formal) is on
https://thebiganswers.wordpress.com/2016/02/03/why-carrier-is-right-on-fine-tuning/
Panpsychist
I can’t follow your notation.
I can just confirm:
P(that we would observe FT|if Life and God exists) is less than 1.
P(that we would observe FT|if Life exists and God does not) equals 1.
Whether that gets the inequality P(God exists|if Life exists and FT is observed) is less than P(God does not|if Life exists and FT is observed) depends on the prior probabilities of God exists and God does not exist (and, if that gets tight, how much difference there is between P(FT|L,G) and P(FT|L,~G)).
So, if you assume the priors for God exists and God does not exist are equal, then you get that result. And in TEC I use 1:3 prior odds on God exists, so the inequality holds there, too. But if someone can argue successfully for prior odds favoring God, and by enough, they can reverse the inequality (then, FT is still evidence against God, but not strong enough evidence to conclude he doesn’t exist).
Your discussion of point (4) has me imagining a god floating around in the meta-universe, pointing its hyperfinger this way and that, intoning “Let there be light! Let there be gzorch! Let there be jingblatt! …”
Well, I suppose that’s one subset of possible outcomes. 🙂
But just so no one mistakes that as all I meant, there are many other subsets, e.g. Gods who are content to just exist and don’t want to create anything, Gods who create lots of cool stuff (stuff cool to them at least) but none of it is intelligent life, Gods who only create angels and magical paradises, Gods who create humans but only inside magical paradises, and Gods who can’t create much and so don’t even try, and Gods who aren’t smart enough to create universes like ours.
I’m not even sure that exhausts all possibilities. But it’s a start.
I ended up spending a lot of time reading Barnes’ blog. PZ Myers, Victor Stenger, and Hector Avalos all showed up to defend themselves, but William Lane Craig did not, because (guess what?) he’s not bad. He even recommends Craig’s book “Reasonable Faith: Christian Truth and Apologetics” as the “best resource” when debating him on fine tuning, this coming from a self-identified expert on fine tuning. So it’s the atheists who are wrong on fine tuning according to the list of people he chooses to engage. No other Christians are called to the carpet. Either other Christians are so far off the mark as to not even warrant mention or their work on fine tuning is impeccable, but either way he made up his mind that fine tuning is the one thing that proves God for him.
Even as this discussion goes on, the underlying point, which is that the fine tuning argument appears superficially convenient for Christians, but really weighs in heavily against them, is inconvenient to Barnes so he would like to avoid discussing it, so he focusses on nitpicking. He felt the need to cite Tim McGrew’s bad joke. This suggests to me that undermining your credibility is more important to him, also judging by those he chooses to critique.
I came away with no more deeper sense that Christians have an unacknowledged psychological need to make it seem like we’re incompetent fools because the actual points against them are just too frightening. I had quite a number of thoughts, but they all ultimately boiled down to that one simple thing.
Christianity makes a vast number of ridiculous claims, a mountain of a house of cards, all of which are under attack, all have actually been knocked down. But any individual Christian only engages with one at a time, and all the other knocked down cards automatically get put back up in their minds when they move to a new topic. What Barnes is doing is what all Christians do. It’s unprofessional, but it can look very professional, but not in this case because there are too many insults coming from both of you.
No amount of math is going to rescue any specific claim of Christianity. The likes of Barnes are playing a social game by pretending that some little math error indirectly validates their beliefs. Every little battle won is a huge victory for them, they can put back up one card, but they’ve preemptively put back all of them and remade the whole house of cards based on one nitpick. Yet no one comes along to a blog like Barnes’ and thinks the Resurrection is true because he claims someone made a math error on a completely different topic or more likely a semantic one. It’s a psychological predisposition to feel the whole house of cards is a solid rock based on one little nit which may not even be accurate.
They have to keep this inane discussion going because Christianity is not dead, it’s the majority opinion dying the slowest of slow and painful death because of that sickest of Christian Sins, Pride. This is why arrogance is their most devastating and frequently leveled insult.
That’s an interesting observation. Because I believe that’s the case more broadly, i.e. that these debates often become really about that. And sometimes that’s legitimate. When you are being honest and valuing accuracy, discrediting authorities is the correct and most effective way to free people from their influence. But like any weapon, this can be used for evil as well as good. Iron makes both pots and swords. And so you see creationists and antoi-vaxxers doing this all the time: ignoring most of the actual debate over evidence and logic, and instead trying to fabricate attacks on the credibility of those opposing them. In many cases this is deliberate dishonesty. But I also think in many cases it’s delusional: the people who do this, don’t realize it’s what they are doing, that this is their real motive, and that it has caused them to ignore evidence and arguments that refute them, and compels them to try and find evidence their opponent’s must be wrong, which their delusion then leads them to see, even when that evidence doesn’t really exist.
From Tige Gibson:
—
Luke Barnes does not seem to ever come out cleanly in support of his god, which I find intensely irritating, because it means that he’s hiding behind the respectability of his profession (cosmology) to drive what is essentially an apologetic agenda. My interest in him is primarily associated with this deception because I am willing, for the sake of the argument, to give him (and any other god-botherer) deism – which is all that fine-tuning will get him, even if he is completely successful with any “fine-tuning therefore god” argument and deism is functionally equivalent to atheism (because a god that does nothing sounds a heck of a lot like no god).
There’s also the fact that, if there is in fact “fine-tuning”, then this tells us something about the fundamental nature of the universe and our efforts should be dedicated to investigating that rather than dealing with the superstitious rabble who lurk around the edges of science trying to justify their god. A very irritating waste of time.
Note also that Barnes has done worse than recommending WLC’s book, he shared a podium with him at the 2015 St Thomas Summer Seminar as the other pro fine-tuning speaker (go to the bottom of the page and open the Previous Seminars drop-down box to see the 2015 line-up). I’ve yet to see any footage of that particular seminar, but I am keeping an eye out for it.
Richard, I think it’s crystal clear from Luke’s posts that he does not think “you can’t put hypothetical possibilities in a reference class, then he must conclude the fine tuning argument invalid.” In fact he’s pretty clear he thinks the EXACT opposite of that. Rather, he’s saying that’s YOUR position, because you are claiming probability only measures the frequency of things happening or of things being true. Here’s the paragraph:
“Carrier’s approach to probability is inconsistent. He keeps shifting the goalposts. In TEC, when talking about a cosmic designer, he says “Probability measures frequency (whether of things happening or of things being true)”. Only known cases, verified by science, can be allowed in a reference class. But now, in OBR, it’s OK to put hypothetical possibilities in a reference class.”
I don’t know how anyone can read that paragraph and not see he is clearly characterizing your approach to probability and how it is contradictory–how you have read that to be him describing his own approach is beyond me. Now he may be wrong that your approach to probability commits you to that, but just on a simple grammatical level you have obviously misread him. In fact you have misread him so badly I really struggle to understand how that misreading can be anything but deliberate. The most charitable interpretation I can think of is you just read him so fast (and angrily) that you made a clumsy mistake. But, I have to admit, I’m really finding it hard to believe that. This is not a small or incidental point Barnes makes in passing–he consistently stresses that he rejects the frequentist interpretation because (among other reasons) it would make cosmology impossible, and that is precisely because it would rule out counting hypothetical possibilities.
I invite Luke to come and correct me if I’ve misstated his position here at all.
That being said, it is, however, impressive that you took the time to respond to him at all, and I think we’re all grateful for your time in doing so. But really, take a deep breath next time before responding.
If I understand you correctly, you are saying he thinks I said “Only known cases, verified by science, can be allowed in a reference class,” and therefore I can’t go against that statement, although he himself disagrees with it. If that’s the case, then as I also go on to point out, I never said that. So once again, he is arguing against a phantom.
But that is confusing. Because if he thinks “Only known cases, verified by science, can be allowed in a reference class” is false, why would he say I am supposed to include non-verified supernatural events as verified when calculating the prior? He appears to argue that if I allow hypothetical elements in a set, I am obligated to do that. And that makes no sense. As I noted.
I am having a hard time getting any coherent argument out of what Barnes writes.
Speaking of publishing too quickly, that should read: … “I think it’s crystal clear from Luke’s posts that he does not think “you can’t put hypothetical possibilities in a reference class.” Full stop.
I couldn’t figure out what correction that amounts to (otherwise I would have just carried it out for you). Sorry! I’ll just leave this up in case others can.
Hi Richard,
I am struggling a bit to follow your first argument and connect it to mine (Notice i used H and not “observations”).
How do you define “Observations” in what you derived? Which of the following are instances of “Observations”?:
A: “It was observed by a human that earth has an atmosphere with oxygen”
B: “Earth has an atmosphere with oxygen”
C: “Bill Clinton was president in the united states”
I am simply trying to understand if Observations and a general hypothesis H (which is what I wrote) are the same thing on your view.
Secondly, i am trying to translate:
“P(observers) is always 1 when anything is observed.”
into a statement about probabilities. Firstly, when we are talking about the event “observers”, we need to distinguish between *any* observer (i.e. there is at least one thing in the universe doing observations) and specific instance of observers which we know exist (i.e. human life, including you and me). I thought the argument was about the later which is actually the information we have available.
Now more importantly I am trying to translate the statement into probabilities (the “when …” part). do you mean:
“P(observers|something is observed) = 1”?
which is of course true as a matter of semantic but I am failing to connect this to our actual discussion about priors.
As to the lottery example: I am pretty sure I agree and have agreed all along.
As to the priors, well, I agree coming up with priors for naturalism or the existence of a God is nearly impossible (see my document, the last section), but that is what you do in TEC (i.e. footnote 8) and it is the argument around those priors I am examining.
Is the argument in footnote 8 for a prior of 1/4 about p(G|Ob) or p(G|b)? (O here means observers in the second sense, i.e. you and me and all other humans). I.e. which of the following is true:
p(G|b) = 1/4?
p(G|Ob) = 1/4?
p(G|b) = p(G|Ob)?
No.
The hypotheses are and are only “gods made everything (universe + observers)” and “no gods made anything (universes + observers).” I have never put “there are observers” as a hypothesis. Or “there are observations.” Those are always in either e or b, not h.
When I state merely P(observers), I’m leaving unanswered “given what,” e.g. gods or physics. My next section then addresses that. So, if it helps, read it as P(observers|gods,~gods).
Or just ignore that and focus on the point that actually matters:
The next question you want to ask is, putting observers in e, is there a difference between P(observers|gods) and P(observers|~gods), before we input fine tuning? Problem number one. Then, we have a posterior probability of gods vs. ~gods, which becomes the prior when when we put fine tuning into e. And at that point observers has moved to b (method of iteration: once you get a posterior from e, e goes into b for the next run of the equation, when you add a new item of e). From there I think you see what I am, that fine tuning is 100% expected on a b with observers and h = no gods (and so observers who already exist, observing fine tuning can never increase the probability of gods). So I’m assuming you want to tinker around somewhere before that happens. That’s what the rest of my comment pertains to. How do you get “a posterior probability of gods vs. ~gods” using only observers? Where do your priors come from? Where even do your likelihoods come from?
The priors are a problem. But maybe we can just principle of indifference it and say it’s 50/50. I won’t question that for the present case.
But the likehoods in this case are also a problem.
We can grant for argument that maybe we can get a likelihood from ~gods if we assume fine tuning requirements (and their consequence in terms of probability) can be deduced from h as a logically necessary consequence of h (I’m assuming you do not mean to just put fine tuning into b or all of observed physics for that matter!). I’ve noted it’s actually doubtful we can get a number for that, actually, even if its deduction from h is granted (and one could question even that). But like I said, let’s assume all these problems have magically vanished, and we somehow actually can deduce P(observers|~god) using fine tuning as an entailed consequence of ~god.
What then is P(observers|god)?
We don’t know. We don’t know how lucky we need to get in choosing a god, to get observers at all.
It gets worse when you have to narrow that god further to “predict” not just observers, but all the other data (e.g. a universe that looks exactly like one would have to look if there are no gods). But even setting that aside, we still can’t calculate P(observers|god).
So there is no hope to this procedure.
Even if we could calculate P(observers|~god). Which I’m also explaining (separately) we can’t do either.
If instead, you don’t iterate but just throw all the evidence in e, getting e.g. P(observers+finetuning+all other observationa|gods) and P(observers+finetuning+all other observationa|~gods), you still need the probability that gods will make observers (and the prior probability of gods vs. ~gods), and thus how lucky a choice of god we need for that, which we don’t know. So you can’t get anywhere that way either.
Since the procedure can’t be performed, this approach cannot be used.
Which is why I don’t use it.
True.
There could have been different kinds of observers (e.g. angels in a paradise).
That’s just not what we observe.
If we need to explain why we are the particular kind of observers we are, now we have to get an even luckier choice of god.
And our subsequent observations then rule out (i.e. greatly reduce the probability of) other observer-making gods (who would only make, e.g., angels in paradises).
You can try moving this data all around in the equation. It doesn’t help. You still don’t know even P(any observers|gods). Much less the prior probability of gods vs. ~gods. And taking the data out of the equation gets you a result not pertinent to our world…until you put that data back in, and then you get a reduction of the probability of god (because all our observations are less expected on gods than ~gods).
Yes.
The connection might not be relevant. If you understand that observations entail observers, and that statements about what would be true of unobserved worlds are not statements about our world, then we can move to the question that matters: how you determine P(observers|gods).
Not really. You haven’t discussed my method of determining the prior there at all. So I have been assuming that isn’t what we are talking about.
You are the one who wants to change the methods to something else. My methods make sense for what I do in TEC. But you want to do something different. It’s “that something different” that we are here talking about.
And that has mostly consisted of your trying to move observers from b to e, which on first analysis is illogical (because b would then lack observers, rendering the equation irrelevant to us; any conclusion predicated on their being no observers is of no relevance to observers). But I’ve been trying really hard to operate within your framework, where it makes sense to ask what the probability of observers is given that there are no observers. The only way to do that (?) is to pack the potential for observers into the hypothesis: gods or ~gods. But that then requires you to know what you don’t: what P(observers|gods) is.
(Again, it also requires knowing what P(observers|~gods) is and I’ve been pointing out why you and Barnes don’t really know that either. But again, that’s a separate argument.)
Remember, in TEC, O is an undeniable, so it goes in b. It is background knowledge. “Given that there are observers, what do we expect to see?” The answer on ~gods is exactly what we see. The answer on gods is not what we see. That generates a likelihood ratio favoring ~gods.
That’s the argument in TEC.
You are the one trying to argue, instead, that O should be removed from b. So here, lately, we’ve been discussing that.
If you do that, then how you determine the priors is not what I do in TEC (where I assume not just O but all past established knowledge is in b, e.g. our hithereto observed frequency of supernatural agency vs. physical causes of ordered events; which is normally what you put in b). You have to do something else.
And that’s what we are discussing now. Your argument. Not mine.
You are the one who wants to take O out of b. Which eliminates all contents of b (except logically necessary truths). All physics, all past observations, everything. Because those can’t exist without O. Any set that includes those, includes O. They are logically inseparable.
That has some very strange consequences.
And that’s what I’m trying to explain to you.
RE: Footnote 8
Note: at no time will I argue we should get rid of O. In fact I will demonstrate that it is Richard who has got rid of O, in Footnote 8.
Preliminaries:
Sometimes the letters we use in the Bayes formula don’t stand for atomic propositions but conjunctions of propositions. For example we can define B := OXYZ. It is just, often, more convenient to use one letter instead of a long string of them.
If we want to we can define B’ := XYZ, in which case we can say B = OB’.
Given this, if we have a prior probability, say, that is p(H|B) we can write it as p(H|OB’). We haven’t changed anything, we are just using different notation for the same thing. Why do it? For the argument I will make it is just to make explicit that we are supposed to be conditioning on O.
What is a conditional probability? In terms of our notation, we are supposed to reduce our total population to just those items that meet the condition on the right side of the pipe, |, and then divide the count of the things on the left by this number. So in p(H|OB’) we reduce the total population to those items that meet the criteria OB’ and see what fraction of those things have H as true.
Footnote 8:
We are told to start with nothing in our background. Clearly this means some universes in our population will not have life. Logic alone does not suggest life exists in all universes.
We divide this in half, giving half to god and half to naturalism. Dividing a population in two does not get rid of anything so obviously non-life-bearing universes are still in our total population, though we haven’t said how they are distributed among the halves.
We divide the god half in half again to get the intelligently designed universes. Dividing in half doesn’t change the fact that our total population still has non-life-bearing universes.
We are told to use 25%, which only makes sense if we are dividing the intelligently designed piece by the total population. But, again, the total population contains non-life-bearing universes.
So, before dividing, have we reduced our population to only those that meet the criteria on the right side of the pipe? No. So have we actually calculated p(H|OB’) (and therefore p(H|B) ) ? No, we have not.
In order to do so we would need to know the fraction of naturalistic universes that are life-bearing, add it to the god fraction that are life bearing, and then use that as our denominator. e.g. 0.5 (0.5) / (0.5(0.5) + 0.5 (p(O|N) ).
If your position is that we can’t know that number, or that we can’t know that it is different from the number we used for god, then you should set p(O|N) to 0.5 as well. In that case we would get 0.5 (0.5) / (0.5(0.5) + 0.5 (0.5 ) = 0.5.
So the number arrived at in footnote 8 is the wrong number to use in the Ikeda and Jefferys argument. In that argument B is supposed to include the proposition O.
Einniv, I think you are confused.
Note 8 is not in the section about fine tuning. It is several sections earlier where I talk about the prior probability of gods.
And note 8 does not say I scrubbed b of content. I didn’t. In the text I am very clear it contains all actual established background knowledge of the human race. Note 8 says, rather, that even if I did scrub b, we would still get the same result. And then I show how (by nothing other than a purely logical division of the possibility space). The result does indeed exclude O from b! But in the text, I do not use this method. I used a b full of our actual knowledge there. Note 8 is just a digression for anyone (people other than me) who want to try getting a prior with an empty b.
Finally, that logical method splits the space evenly between godless and godfilled possible worlds (by principle of indifference on the binary proposition), and then splits the god half between designing and non-designing gods (again by principle of indifference on the binary proposition). It thus includes the absence of observers—at least one quarter of the possibility space in fact is dedicated to worlds that never contain observers (except God). And observers is thus not in b. It is here a component of h. How the other half gets divided then depends on how one ramifies that space (e.g. note 20 argues it does not divide the same way; although to gainsay that, one would have to argue the point—and do so with only a priori true propositions).
But yes, if you choose to split the godless side evenly as well, you get a priors distribution:
P(God&observers) = 0.25
P(God&~observers) = 0.25
P(~God&observers) = 0.25
P(~God&~observers) = 0.25
This then generates a consequents distribution on e = observers:
P(observers|God&observers) = 1
P(observers|God&~observers) = 0
P(observers|~God&observers) = 1
P(observers|~God&~observers) = 0
This then gives us a posterior on zero information but that one e (plus all logically necessary truths):
P(God|observers) = 0.50
P(~God|observers) = 0.50
This then becomes the prior probability when looking at the next items of evidence, by which point “observers” then is in b. As happens when you use the posterior of one run as the prior in the next (that procedure ports everything in e in the first equation into b in the second). This is standard iteration (see Proving History, index, “iteration, method of”).
So a case could be made that my note 8 should reflect this result instead (absent a successful argument to the contrary, e.g. the one in note 20). But since I don’t use that method in the text, that’s moot to what we are discussing presently.
“What Barnes doesn’t get is that fine tuning will be observed in all observed atheist universes”
I don’t get this. First of all, he cannot get what is not true. There could be plenty of observed universes for us, which are not atheist, yet which do not exhibit a high degree of fine tuning. For example, if our universe is a simulation, it is still our universe for these purposes.
But if the statement is modified to “What Barnes doesn’t get is that fine tuning will be observed-from-within in all observed-from-within, designed, universes” or something like that, then surely Barnes gets this. It sure seems to me from reading his posts that he gets this.
But his point seems obvious to me. One has to look at the non-observed universes as well, in which no observers came into existence. Maybe Carrier thinks otherwise, but then this is where Barnes is wrong, not that “he doesn’t get” that fine tuning will be observed-from-within in all observed-from-within universes (actually, one has to add the proviso that the observers have the capacity to make this observation, and actually attempts to do so),
That may be. I’m trying again in this thread here.
But if all observed undesigned universes are fine tuned, fine tuning as evidence cannot increase the posterior probability of design. That does not seem to be registering for Barnes. To use it as evidence of design that way, you have to have a P(e|design) that is higher than P(e|~design) for e = fine tuning. And that is logically impossible. (As even Collins conceded.)
Barnes I think wants to somehow get fine tuning to lower the prior probability of undesigned observers existing, rather than the consequent probability that they would then observe fine tuning (which is always 1). And that’s where the weeds are.
I know that academic papers are often rather impenetrable to lay readers. It’s probably for that reason that I set off on arguing something similar to Ikeda and Jefferys and later yourself, but I think that my approach might be slightly more accessible.
—
I’d be interested in your comments. Here’s how I tackled it. Note that in this instance I was focusing on William Lane Craig’s argument for god via the resurrection, but the principles remain the same when someone like Barnes is trying to use the appearance of fine-tuning to argue for a cloaked god (a non-terrestrial intelligent designer or whatever wording any theist wants use to obscure their intention, rarely will they come out and claim god if they want to maintain a veneer of scientific respectability).
Note also that I ended up (by using what I term “Plantigerant probabillity”) with a likelihood of the resurrection being true of 0.5 – this is ridiculously high, but even so, it doesn’t help the theist’s argument. They can only help themselves if they increase the likelihood of the resurrection, which they will only do if they are predisposed to the god conclusion. This goes to show how the apparently objective processes of Bayesian probability can be fiddled to get the result you want, if you are intellectually dishonest enough.
Sorry I won’t have time to review that. But thanks for the link. Others among my readers might explore it and comment there.
Thanks for replying, Richard.
“P(that we would observe FT|if Life and God exists) is less than 1.
P(that we would observe FT|if Life exists and God does not) equals 1.
Whether that gets the inequality P(God exists|if Life exists and FT is observed) is less than P(God does not|if Life exists and FT is observed) depends on the prior probabilities of God exists and God does not exist”
Yes. But these two do get us the inequality that P(God exists|life exists and FT is observed)<P(God exists|life exists). If thought this was your main point and a major source of your frustration with Barnes, e.g. in "What Barnes doesn’t get is that fine tuning will be observed in all observed atheist universes”.
Panpsychist
I’m still not certain what you mean. P(God exists|life exists and FT is observed) could still exceed P(God exists|life exists) if the prior probability of God is especially high. So I assume you mean if that prior is not especially high, then FT lowers the probability of God even more (rather than the other way around as FT advocates claim). Or, of course, that if the prior for God is 0.50, then FT pushes the posterior probability of God below 0.50.
“That may be.” I assume the word “that” refers to the proposition which is the negation of the proposition Carrier has being asserting, viz., “What Barnes doesn’t get is that fine tuning will be observed in all observed atheist universes”.
So Carrier maybe is now finally agreeing that Barnes *does* get that fine-tuning will be observed in atheist universes. I guess that is progress. I think Barnes should be given credit for getting such basic things. He seems to be an expert in this area after all (we should also note that he believes P(Fine Tuning| God) = P(Fine Tuning| not God) = 1).
“Barnes I think wants to somehow get fine tuning to lower the prior probability of undesigned observers existing, ” I don’t know, but I would hope so, because it seems kind of obvious that this is true. I don’t see what the brouhaha is about.
Suppose roughly four competing hypothesis, LifeFriendly, Multiverse, AnythingButGod, and BigFluke where LifeFriendly is “there is some life friendly force involved in the universe coming into being as it is” — these are to be defined to be exhaustive. But, given also a strong belief that the form of the current physical theory is basically right, but that there is also no reason in that theory why the constants are the way they are, to be in the very small region that makes life possible, it seems obvious that their being in that region raises the posterior probability of LifeFriendly substantially, since P(Nice constants in actual universe| LifeFriendly) would be much higher than P(Nice constants in actual universe | not LifeFriendly). I have seen some people try to put “Life exists” into the background conditions including apparently Carrier, but this seems like a distorting selection effect which amounts to an egregious begging of the question.
Now, of course, if one’s prior for “Life friendly” is very low to begin with, then one of the other ones might be a better bet. Many people have a very large prior for the AnythingButGod (aka NoWooPlease, etc) hypothesis. And there is some independent evidence for Multiverse, although perhaps not for a Multiverse big enough to get a largish P(Nice constants in (some part of the actual) universe | Multiverse).
You need to read my argument in TEC, in particular note 31. We have already refuted what you are saying and moved past it. You are weighing in too late in the conversation.
We actually don’t know the size of the life friendly portion of the possibility space, in either the God or Godless portions of the possibility space. I have explained why in this very article above. You seem not to have read it; because you don’t seem to be aware of those arguments and aren’t responding to them.
But we also don’t know what the threshold probability is, i.e. how locally improbable something has to be for its total probability to be low. For example, winning a lottery has a low probability locally, but that a lottery will be won has a very high probability in total (because many people are playing, therefore the odds of a winner are vastly higher than the odds of a specific winner). The same goes for fine tuning: even if we could establish how improbable fine tuning is (and we can’t, per above), that’s still only a local probability. We don’t know the total probability. So, for example, if the universe will last a trillion years, the probability that at least one event in it of local probability 1 in 10^150 will occur is in fact very near 100%. That event could as likely be its origin as anything else. So 1 in 10^150 is not all that improbable. Undesigned events of that improbability have already happened and will continue to happen. It’s therefore not an unlikely event (speaking cosmically).
What we really want to know is the relative probability of that happening vs. getting a correspondingly lucky God. That is, maybe the right kind of God has a probability of 1 in 10^150. And if the probability of the same result without a God is 1 in 10^150, the relative probability on this evidence alone is 50/50, i.e. there is a 50% chance that we got the 1 in 10^150 lucky godless universe and a 50% chance we got the 1 in 10^150 lucky God.
But we can’t work that up, because we have no sufficient information to calculate any of these probabilities.
Michael Ikeda and Bill Jefferys are committing three fallacies in “The Anthropic Principle Does Not Support Supernaturalism”: the obfuscation fallacy, begging the question and the ambiguity fallacy. Their argument goes as follows:
1) Fine-tuning says: There is an astronomically small chance P(F) that life emerged naturalistically
2a) We observe that life emerged naturalistically (begging the question = circular reasoning)
2b) We observe that life emerged (ambiguity = two definitions of life friendly)
3) P(F|N) differs from P(N|F) (obfuscation = irrelevant theory)
4) From 1) and 2) it follows that P(F) = 1
5) The smaller P(F) is claimed to be, the more the fine-tuning argument is false.
This is a text-book example of fallacious reasoning. The argument fails to make any conclusions about supernaturalism or why 2a would be true.
That’s not correct. That isn’t how they argue at all.
They are arguing “if fine tuning is a fact, then the probability of supernaturalism is reduced, not increased.” You don’t even have that right. Moreover, their argument follows on the simple basis that there are more ways to get life if supernaturalism, whereas fine tuning is the only way to get life if ~supernaturalism. The conditional is logically necessarily true. And is simply an argument that fine tuning is evidence against supernaturalism. Not that it refutes supernaturalism. Nowhere in their paper do they say that.
So you do not seem to be reading their work with any care. You’ve certainly not described it correctly.
And I should remind you, Sober independently corroborated their results (without their being mutually aware).
If I reconstruct it as a Bayesian argument, then it looks even worse. Only statement 4) will be modified:
1) Fine-tuning says: There is an astronomically small chance P(F) that life emerged naturalistically
2a) We observe that life emerged naturalistically (P(F) = 1) (begging the question = circular reasoning)
2b) We observe that life emerged (ambiguity = two definitions of life friendly)
3) P(F|N) differs from P(N|F) (obfuscation = irrelevant theory)
4) From 1) and 2) it follows that P(F) can be very close to 1
5) The smaller P(F) is claimed to be, the more the fine-tuning argument is false.
Why would you weaken an argument if it follows deductively from the assumptions? Can you provide an equally clear reconstructed argument with true statements that follow logically from each other?
I know you have a popularized version. Would you endorse my reconstruction of it:
1) Life-containing universes with supernatural laws do not need to be fine-tuned
2a) Life-containing universes with supernatural laws are never fine-tuned
2b) Life-containing universes with supernatural laws are not always fine-tuned
3) When a life-containing universe is fine-tuned, it does not have supernatural laws
I would say that the transition from 1 to 2a is invalid, just like that from 2b to 3. So again, I cannot reconstruct a valid, sound argument.
You say there are no ways to get life without fine-tuning and without supernaturalism. I agree. But this results in a clear proof that, given fine-tuning, the chances of supernaturalism are increased, because supernaturalism is the only paradigm that can create the fine-tuning systematically. The basic assumption in the whole discussion is that naturalism can only depend on mere chance to create the fine-tuning.
(A) That isn’t their argument. I’ve told you this. You need to attend to what they actually argue.
(B) Fine tuning does not say “there is an astronomically small chance that life emerged naturalistically.” Fine tuning says there would be, but for fine tuning. And it is fine tuning that supposedly has an astronomically small chance, but for intelligent intervention.
(C) They make no assumptions about whether intelligence intervened to produce life after fine tuning. Their argument is solely about the effect of fine tuning on the probability of supernaturalism (and all they conclude is that it lowers rather than raises it). The effect of other facts is not even addressed.
(D) There is not an astronomically small chance that life emerged naturalistically. That has nothing to do with their paper (they make no assertions about it, nor is it a premise in any of their arguments). I mention it here only to remind you. I’ve told you this already. I’ve directed you to peer reviewed literature.
Since you aren’t describing any argument they made in their paper, you are not addressing anything that exists. Your remarks are irrelevant and impertinent. And suggest you are stubbornly uninterested in the truth.
Peer reviewed arguments do not need to valid in order to become cited often. For arguments that are not peer reviewed, however, being valid is the only way to become cited often. Therefore being peer reviewed is always evidence against the validity of the argument.
My apologies for the parody, but this is how, in my opinion, your popularized argument works.
Best regards,
Ward
That’s insane.