A reader pointed something out to me that was a fantastic facepalm moment. It’s another demonstration of how Bart Ehrman doesn’t know how epistemic probability works, and not only hasn’t read On the Historicity of Jesus, he doesn’t even know what it argues. This leads me to two general lessons I hope my audience has already learned, but that he certainly needs to learn. The first is that everything that isn’t logically impossible always has a nonzero epistemic probability of being true. And recognizing this is fundamentally essential to all sciences and knowledge-seeking fields. The second is that if you want to challenge an assigned probability, you have to first understand what the claimant is measuring—what are they saying their assigned probability is a probability of? And recognizing this is fundamentally essential to all sciences and knowledge-seeking fields.
I’ll explain these two general lessons using a recent example of Bart Ehrman failing to do his job as a historian. I’ve noted recently that often enough Ehrman does not read the peer reviewed literature on the subjects he has an emotional investment in—not just mine (which is bad enough, since my book is not only the latest but the only peer reviewed book ever published on the question of the historicity of Jesus in nearly a hundred years), but even articles I cite from the peer reviewed literature that support me (likewise the peer reviewed literature that supports others he is intent on disagreeing with, from Murdock to Doherty to Goodacre). And in result he argues against things I didn’t say, and then he doesn’t argue against established peer reviewed arguments to the contrary of his position. I also noted that he plays fast and loose with the facts, but more bizarrely, he insists that history has to be about probability while declaring probability theory inapplicable to history. I’ve said before that this is because he doesn’t know how probability works. Now we have an example. (And hat tip to Josh for calling my attention to it).
On Ehrman’s blog, a reader posted a remark on November 6 (2016) that said, among other things:
[….] Speaking of proving Carrier wrong with mathematical precision, being a social scientist who regularly works with mathematical models, one of the things about his argument that I find insulting is his attempted use of math—specifically Bayesian statistics—to prove the improbability of an historical Jesus. His method is so error-ridden that he should be embarrassed to even express it in public, let alone actually publish it as a work of scholarship. For starters, Carrier starts off by assigning probabilities, not to the most mundane aspects of a living person, such as name, gender, nationality, time and place of birth, etc., but instead to the most conveniently ridiculous claims, such as virgin birth, resurrection, miracles, etc.
That’s like if I were to attempt to prove that the Buddha Siddhartha Gautama never existed by assigning probabilities to his sitting under a Bodhi tree for forty days, being born from his mother’s side (instead of through the birth canal), and eventually achieving Nirvana after death. I would have to pull those probablities out of the same place that Carrier gets his probabilities for Jesus; namely, ex rectum. I mean, let’s be serious. How does one find the probability of parthenogenesis? The best we can say is that the probably is either 0 or 1; that is, either it’s totally impossible or it’s possible, because if it’s determined to be possible (via observation or experimentation or whatnot) then it becomes plausible, and once it become plausible then it becomes probable.
And that’s why trying to assign probablities to miracles is a fool’s errand. Once you assign it a probability above zero, you’ve automatically made it “possible,” and if it’s “possible” then it’s not a “miracle”. That’s what logicians call a category error.
The next day Ehrman responded:
Thanks—that’s a lucid explanation of the problem of his use of Bayes statistics.
This is the facepalm moment. Because everything that reader said is false. It’s false even on basic probability theory, and yet he or she claims to be an expert in that because they are a “social scientist.” But it’s also false on the basic facts: I have never “assigned probabilities” in this debate to miracles actually happening. That is nowhere any part of my book—where in fact, exactly to the contrary, I declare I will not even consider models of the historical Jesus that propose any miracles, that I will only consider miracle-free historical Jesus models, and indeed only the most credible of them (see OHJ, pp. 14, 26, and 30, where I declare any miracle-reliant theory of historicity ignored for the remainder of the book). The only time I ever even “assign probabilities” to anything connected with miracles, I explicitly deal only with the frequency with which such stories are invented (not “actually happen”), for real persons vs. invented ones (OHJ, Ch. 6). And I am so explicit about that, there is no possible way anyone who read my book could mistake the matter. So that means this reader never read my book. It also means Ehrman never has, either. Because he actually thinks this commenter is correctly describing what’s in it!
Ehrman’s remark here is revealing in a way very similar to what happened in the Penis-Nosed Statue case, where Ehrman had remarked on his blog at how good and useful some research was that some fans posted there. Then their claims were shown to be shoddy and false, thus exposing the fact that Ehrman didn’t know that. Which means he never did any of that research himself, and thus actually didn’t meet even the most rudimentary moral responsibility of a historian when responding to Dorothy Murdock on the matter of the Penis-Nosed Statue—and then when he was caught, he tried lying about it instead of admitting he screwed up. (For summary, links, and evidence on that case, see my Ehrman Recap, Item 11.) Here he has done this again: with a single remark Ehrman has proved he doesn’t know what this reader is saying is false—both as to probability theory and as regards what I actually argue in OHJ. Such a man has no competence to debate this matter. No opinion he has of my work is valid, if he neither knows what I argue, nor how probability works.
A good analogy exists with Thomas Thompson’s case against the historicity of Moses—a conclusion now generally accepted by the mainstream consensus, despite having been fought by the same kind of uninformed and reactionary tactics used or endorsed by Ehrman on this question now. (See my summary of the Davies article on this whole point.) Imagine if a reader had posted on Ehrman’s blog that Thompson was an idiot because he was assigning probabilities to Moses actually parting the Red Sea and then from that concluding Moses didn’t exist (“What a fool! Hahhaha!”). Ehrman would probably not say “That’s a lucid explanation of why Thompson is wrong.” No, he would correct his reader by pointing out that that’s not what Thompson did at all, that Thompson doesn’t assume the only options are miracle Moses or no Moses, and that Thompson’s case has always been against a mundane Moses, not a miraculous one. (And that Thompson’s resulting conclusion is widely accepted—even by Ehrman himself.)
Of course, if Ehrman also understood probability theory, he’d also have corrected this reader’s atrocious epistemological logic. But I’ll get to that next. First, let’s explore the more important matter of what I (like Thompson) actually argued…
Taking the Second Point First: Know What Your Opponent Is Measuring
I nowhere assign probabilities to miracles in OHJ. Ehrman and this reader evidently have not read Chapter 6 of OHJ, where I explain what I am assigning the probabilities to, and the data I am deriving those probabilities from (in respect to where anything about a miracle comes up): the frequency not of such events (like parthenogenesis) happening, but the frequency of inventing so many details like that for a real man rather than a fictional one; and I derive that frequency from a long list of actual prior examples of that being done—to both real and fictional persons. A real social scientist would recognize that procedure as valid; it’s exactly what they would do. The only difference is that my database is small (relative to what social scientists deem adequate), so I need very wide margins of error to maintain validity, and indeed that’s exactly what I generate. If he or she wants to argue those margins should be wider, they need to make such an argument from the data, not the armchair. But it’s clear they don’t even know what my margins of error are, or that I even assigned any.
Nor do they know what I am measuring. That reader essentially is lying—insofar as he (or she) is representing himself (or herself) as actually knowing what I argue. If they actually knew, they would know that I am not measuring anything as absurd as “the frequency of parthenogenesis.” Rather, I am measuring the frequency of “inventing for a hero a myth of parthenogenesis.” And not merely that, but the frequency of inventing a huge array of mythical claims—not just that one; and not just any mythical claims, but a well-established popular homeostatic cluster of them. And not merely that, but the frequency of that being done for historical relative to nonhistorical persons. There is nothing here about measuring how often those absurd claims are true—of anyone. The only thing I’m measuring is how often such a set of claims gets generated for someone (and those claims being true or false is completely irrelevant to that question).
And as it happens, when you look at all past precedents of that happening (of which we have at least fourteen from the relevant causal historical context; fifteen counting Jesus, for whom it indisputably also happened), not a single one happened to a historical person. Fourteen cases. Not one historical. That’s a trend. And it entails that anyone to whom that same process happens, is not likely to be historical (OHJ, pp. 231-32). Historicity is thus not what the evidence of past cases indicates. It indicates that this particular process only happens to nonhistorical heroes. One can then say that maybe Jesus is the exception, the singular bizarre exception out of all known fifteen cases. And that’s true. And you can calculate a probability for that, and I did (OHJ, pp. 242-43; cf. pp. 239-44, on the Rank-Raglan type as my model). One can then say maybe it’s a fluke somehow, that it’s just a coincidence that no historical persons ever got typed that way, despite it happening fourteen (and now fifteen) times. And that’s true. And you can calculate a probability for that, and I did (ibid.), leaving me with my margins of error (which I classify as a judicantiori and a fortiori: OHJ, pp. 596-99).
So, again: The question I ask in OHJ is not “how frequently do such miraculous events actually happen” (nor do I then conclude the converse of that is “Jesus didn’t exist,” which indeed would be a violation of basic logic) nor is it even “how frequently do such miraculous events get made up,” but rather, “how frequently do such miraculous events get made up for actually historical persons.” I find the actual data supports a conclusion of never. And that’s not “ex rectum” (which is incorrect Latin, BTW; it should be ex recto, but whatever, this guy or gal doesn’t get anything correct, so why should I assume they know Latin?). To the contrary, it’s derived from actual data: at least fourteen known prior cases. To get any result other than “never” requires making assumptions not based in any data. Those assumptions can nevertheless be credible, if they derive from what is entailed by probability theory, namely, the probability it’s an “accident” that all fourteen prior times this happened not once did it happen to a historical person. Like parthenogenesis, we’ve never seen it happen to a historical person. But it’s much more causally plausible than parthenogenesis, if the right chance accidents just happened to have occurred. And accordingly, I account for that. This reader on Ehrman’s blog clearly has no idea I did that. Or even what I was doing at all.
In fact, Jesus meets even more markers for mythical persons than the Rank-Raglan type: he is, unlike most historical persons, a worshiped celestial savior deity (OHJ, pp. 96-108, 230), a dying-and-rising demigod (OHJ, pp. 168-73, 225-29), a revelatory space alien (137-41, 146, 197-206), a prophecy-fulfilling godman (OHJ, pp. 141-43, 230), an aetiological cult figure (OHJ, pp. 8-11, 159-63), and a counter-cultural hero (OHJ, pp. 222-25, 430-31; cf. Proving History, pp. 131-32). But these only reinforce the certainty we derive from the most abundant database we have, which is for the Rank-Raglan type. See Should the Gospels Count More Against Historicity? (Ultimately I find that, apart from what we can determine from and for the Rank-Raglan data, nothing in the Gospels argues for or against historicity: OHJ, pp. 395, 506-09.) One could ask the same of the Epistles, insofar as they establish, as even Ehrman now agrees, that the first Christians from the very beginning typed Jesus, again unlike most historical persons, as an eternal, temporally incarnate archangel living in outer space, and as a standard albeit Judaized “worshiped savior deity” with whom apostles communicated by revelation—though I left that to be counted among the determining evidence (in Chapter 11).
So what I am measuring is how often historical persons get that heavily mythotyped (and indeed that quickly, which should be near impossible for a historical person: OHJ, pp. 248-52), not how often historical persons are born by parthenogenesis, or any such nonsense. That Ehrman doesn’t know that, wholly disqualifies him from having a valid opinion on the merits of my case for ahistoricity. Yet in all the sciences, in all fields of knowledge, understanding what a researcher is measuring is fundamental to being able to evaluate the merits of what they are claiming. A social scientist sure as hell ought to know that. And Ehrman cannot call himself a competent historian if he doesn’t know that either.
And this is again, I have to remind you, just the prior probability I’m talking about. No matter what I get as the prior, evidence can always overcome it. So it does not matter that I find the prior probability of historicity to be 1 in 3, owing to how heavily mythotyped Jesus was compared to all other mythologized historical persons. All you need is a body of evidence that’s four times more likely if he existed than if he didn’t, to wipe that prior out and leave historicity as the more likely conclusion. So we still have to look at the evidence. Of course that doesn’t go well for the historicist, either (OHJ, Chapters 8, 9, 10, and 11). Though it could have gone much better (as it does for Socrates, Tiberius, Alexander the Great, Julius Caesar, Pontius Pilate, or Spartacus—even Herod Agrippa). But this reader (and evidently also Ehrman) doesn’t even know that much about probability theory, or about my argument in OHJ.
Taking the First Point Second: Every Logically Possible Thing Has a Nonzero Epistemic Probability
Now, contrary to this liar on Ehrman’s blog, I did not argue “miracles are improbable, therefore Jesus didn’t exist.” But let’s set that aside now (I’ve already addressed it above), so we can learn a lesson this fool evidently never did, about how epistemic probability works. In other words, what a social scientist sure as fuck should already know.
Anyone who actually knows anything substantive about probability theory—and epistemic probability especially—will have laughed their ass off already at the shocking, galling incompetence of this “social scientist” claiming “miracles” like parthenogenesis must have a probability “of either 0 or 1.” Holy fuckballs. Let’s assume they didn’t just commit the fallacy of foregone probability. Let’s assume they mean—and we’ll stick with the one example—that parthenogenesis is contrary to existing science and “therefore” must have a probability of 0. Only a fool who didn’t know how science worked would say such a thing.
Science would be impossible if we assumed everything that contradicted existing science had a zero probability of occurring. That’s the exact opposite of science: that’s dogma. In Bayesian terms, no matter how good the evidence, even if it gave us a likelihood ratio of 10^500 to 1 in favor of parthenogenesis having occurred, we would still have to conclude it never happens—even after seeing that much evidence for it—if we started with a 0 prior probability of it. Needless to say, no scientific progress would ever be possible if we did that. Therefore, “miracles” like parthenogenesis cannot have a 0 prior. And lo and behold, because real scientists don’t take the advice of this knob on Ehrman’s blog, they have discovered parthenogenesis actually does exist and happens quite a lot. Something they could not have done had they reasoned like he or she did.
Now, yes, we haven’t discovered it in humans, and we can produce a plausible, evidence-based explanation for why. But that does not mean we couldn’t be wrong, that no evidence of it will ever be found, that our causal models are infallible, that we are omniscient. And in fact, we can produce parthenogenesis in humans, using barely space-age technology—we’ve literally done it—so it clearly does not have a probability of 0. The same process we use to accomplish it may have an extremely low probability of occurring in nature, but anyone who confuses “extremely low” with “0” can’t really be a scientist. Likewise there may be ways it can happen unknown to us. Like, for example, sorcery. Or gods. There is nothing about science that entails these have a “zero” probability of existing. Science has accumulated enough data to make their existence extremely unlikely. But again, “extremely unlikely” is not “zero.” And no self-respecting scientist should ever be caught confusing the two.
The fact of the matter is, everything that isn’t logically impossible always has a nonzero epistemic probability of being true (Proving History, pp. 23-26; cf. pp. 107-14 and 266-75). And quite a lot of things are logically possible. Including our being in the Matrix; gods and sorcery of some kind being a thing; even parthenogenesis by quantum mechanical accident. So we have to seriously ask how likely these things are, on present background knowledge. For the purposes of determining something as mundane as the historicity of Jesus, we don’t really need to explore that, since we know a fortiori the requisite miracles are far less likely than even billions to one against (see Proving History, index, “a fortiori, method of” and “miracles”), so no theory of historicity that depends on miracles has any realistic chance of being true. That’s why non-delusional historians only consider non-miraculous theories of historicity. And those are, again, the only theories I consider as competing with mythicism in my analysis in OHJ.
But we don’t come to that conclusion because the probability of miracles is “either 0 or 1.” That’s ridiculous. And fantastically ignorant. It’s also fundamentally anti-scientific. As I noted, if we assumed everything contrary to existing science had a 0 probability of occurring, we could never ever be convinced by any amount of evidence that we were wrong, and all progress in science would cease. That’s how Creationists behave. That’s absolutely not how any real scientist, including a “social scientist,” should ever behave. It’s all the worse that he or she somehow wants to define miracle as “impossible,” such that anything that’s possible is ex definito not a miracle. Which is not at all how the word “miracle” is used by anyone, least of all anyone who believes in miracles.
It’s also a fallacy to say that parthenogenesis can never happen because someone calls it a miracle and “we” define “miracle” as “logically impossible.” Not only because they don’t define “miracle” that way, so what “we” define the word as is wholly irrelevant, but also because you can’t change reality by changing what you call it. You can’t “define” parthenogenesis into being impossible. Words don’t have that magical power. And alas, parthenogenesis has been documented, even in humans. So it isn’t impossible. But even if we didn’t have documentation of it yet, it’s still logically possible, and therefore has a nonzero probability of being true; it all depends on future evidence.
This is the difference between “probability” as some esoteric absolute, and what Bayes’ Theorem measures, which is epistemic probability: the probability that something is true given what we currently know—which means this is a probability that is conditional on information we both have and don’t have; which means information we don’t have can change that probability, and therefore that probability can never be zero. In an esoteric absolute sense everything has a probability of 0 or 1, it either actually is true or actually is false. Your probability of winning the lottery tomorrow (indeed, even if you didn’t buy a ticket; because there is a nonzero probability someone else did and you have it in a coat pocket by some accident or design unknown to you) is indeed literally “either 0 or 1,” because, thanks to causal determinism, it’s a foregone conclusion which it will be.
Everything in your life can be described that way. But that’s useless information. The problem you face is not determinism. The problem you face is not knowing stuff—such as what lottery number will be selected to win, or whether someone snuck a lottery ticket into your coat pocket; or indeed, whether someone’s phoning you to tell you you won, means you actually won. And that’s all a question of epistemic probability. When we have really good information, our epistemic probability converges on a physical probability—absent evidence you already won, the epistemic probability you will win a lottery is nearly the same as the actual physical frequency of your chosen number coming up on a hypothetically endless series of winning lottery number selections; it is only not exactly the same, because it has to be adjusted for the physical frequency of other factors that can affect the outcome, like fraud. Epistemic probability is what you end up with when all those factors are adjusted for (see Proving History, index, “epistemic probability”). And yes, this is what’s really going on even when Bayesians describe probabilities as degrees of belief (Proving History, pp. 265-80).
This even extends to the logically impossible, insofar as we can never be absolutely certain that what we think is logically impossible actually is (see The God Impossible), even when we have widely vetted and published logical proofs of the fact (Proving History, p. 25, with p. 297 n. 5). The exceptions are actually extremely limited—basically, just raw, uninterpreted, present experience (by Cartesian principle). Epistemically, everything else has a nonzero probability of being true or false. That includes magically parting seas and turning sticks into snakes, virgin births, psychic battles with demons, partying with Satan, and flying unaided into outer space. We have to take seriously that the evidence does not warrant our believing these things have a probability of zero; so far as we know, they yet could be true. The evidence does constrain how high that probability can be (there is a probability above which we would not be warranted in saying we know human parthenogenesis is more likely than that); but it also constrains how low it can be (there is a probability below which we would not be warranted in saying we know human parthenogenesis is less likely than that). Any scientist who doesn’t know that, sucks at science.
See Miracles and Historical Method for the full skinny on how historians logically must deal with miracle claims. It’s not like what this fool claims. Who doesn’t know jack about historical method. Or probability theory, apparently.
And that makes two strikes. They and Ehrman get wrong what my book argues, and how probability works. The general takeaway is that when this guy (or gal) on Ehrman’s blog says “his method is so error-ridden that he should be embarrassed to even express it in public, let alone actually publish it as a work of scholarship,” he (or she) is actually talking about himself (or herself).
So, a liar posts on Ehrman’s blog that “Carrier starts off by assigning probabilities, not to the most mundane aspects of a living person, such as name, gender, nationality, time and place of birth, etc., but instead to the most conveniently ridiculous claims, such as virgin birth, resurrection, miracles, etc.,” and Ehrman is so uninformed he actually thinks that lie is true. And then this liar becomes an antiscientific fool and makes a wildly false claim about the prior epistemic probability of paranormal events, and Ehrman is so ignorant of probability theory he doesn’t even know that. This is the guy whose opinion we are heeding on this subject? Please explain why. And please also explain why the only way anyone can rebut my case in OHJ…is by lying?
Apart from “tell the damned truth,” there are two lessons everyone should learn here; please do heed them: (1) You have to get right what someone is measuring when they are making claims about probability before you can criticize those claims. You cannot engage in rational discourse in any field of knowledge if you fail to do this. And (2) everything that’s logically possible has a nonzero epistemic probability of being true; so (a) you have to take seriously what it’s epistemic probability actually could be, and seriously ask what facts you justify that probability with, and (b) science fundamentally requires the axiomatic assumption that anything we are certain about could yet be false, and this absolutely requires us to assign every such thing a prior probability more than zero. You do not know how science works (or history for that matter), if you do not know that.