Extraordinary claims require extraordinary evidence. This is a logically necessary truth. Notorious Christian apologist William Lane Craig tries to deny this. But only by playing word games. Let’s see how this statement actually pans out, and how Craig is being dishonest in denying it.

Craig, of course, is an infamous liar—as in, he regularly traffics in misrepresentation and distortion, and sometimes even outright falsehoods (documentation here, here, here, here, here, here, here, here, here, here, here, here, here, here, here, and here; and that’s just a sample; for video surveys, try these and these, and now this and this). By now I think he’s a con artist, who plays rhetorical and word games with people on purpose, to sell them on Christ—or really, I think in fact, dupe fellow Christians into not defecting (the actual purpose of most apologetics I have come to believe, as I noted to Randal Rauser, and since have elaborated in Addressing the New Christian Apologetics). Because this dishonest behavior has burned his reputation with atheists. They no longer trust him. So his apologetic mission simply isn’t believable to any other audience than Christians still intent on trusting him.

Be that as it may, Craig has often stated the maxim that “extraordinary claims require extraordinary evidence” is false. And has even attempted to “prove” this with Bayes’ Theorem—in ReasonableFaith.org question no. 604, “Do Extraordinary Events Require Extraordinary Evidence?” He’s been taken to task for his dishonest rhetorical games when making this claim in many other places. And I’ve analyzed his dishonesty about this in video interviews too. But here I will focus on this, his most “official” written statement on the point, which I take to be a more reliable record of his beliefs and reasoning than recordings of his thoughts off the cuff.

I’ve already given the correct analysis of this maxim in Proving History (pp. 114-17; cf. pp. 72, 177-78, and 252-55; which, incidentally, was peer reviewed, including by a professor of mathematics). The following is well supplemented and further explained there, so I highly recommend folks read that first if they want to contend with my point. By contrast, how does Craig’s analysis go?

The Question Posed

On that webpage, Craig is asked by “Mauro” how Craig can call this maxim false, and even claim Bayes’ Theorem proves it’s false, when “every single Bayesian analysis (besides yours) that I’ve read about [it] seems to support it.” And they then present him a correct Bayesian analysis (emphasis mine):

On Bayesian terms, an “extraordinary claim” would be one whose prior probability is very low (Pr (h/b) <<< 0.5). It means that in order to render the claim more probable than not, we need strong evidence in order to counterbalance [that], which is just to say that we need “extraordinary evidence” for such a claim. Extraordinary claims require extraordinary evidence in order to overcome the overwhelming prior probability against such claims being true.

I would refine this to replace “very low” with “extraordinarily low” to maintain consistency (and I have omitted a notation error I’ll discuss shortly). Though “very low” priors do also require “very strong” evidence, usually when we say “extraordinary” we mean even lower priors than that, and thus even stronger evidence is needed than that, hence “extraordinarily low” priors require “extraordinarily strong” evidence. But either way the principle is the same: when the prior probability of a claim is really low, the evidence needs to be really good in order to believe it.

This is always in fact an inviolable mathematical relation: the lowness of a claim’s prior probability of being true is exactly proportional to the highness required of the evidence’s likelihood ratio to warrant at all believing it. Hence Mauro correctly explains that “extraordinary evidence” exists on Bayesian terms when that evidence is “[extraordinarily improbable] if the extraordinary event had not occurred.” And this “is equivalent to say[ing] that extraordinary claims require extraordinary evidence in order to overcome the prior improbability of the claim.” So they ask Craig to explain how he comes to an opposite result using the same math.

Craig’s Opening Argument

William Lane Craig then opens by declaring this questioner has been “misled by sceptical thinkers who are trying to misuse Bayes’ Theorem to support their scepticism about miracles.” So Craig is very definitely declaring what was just presented by the questioner is a “misuse” of Bayes’ Theorem. I cannot believe Craig honestly believes that; I must conclude he’s the one here deliberately misrepresenting facts and logic to evade an important principle posing a danger to his mission. But if you wish to insist on his honesty, you are left with only one remaining explanation: Craig is grossly incompetent in basic math and reasoning. So is Craig here a liar? Or a victim of the Dunning-Kruger effect? I’ll let you decide.

Craig correctly states and describes the Odds Form of Bayes’ Theorem, as calculating the ratio between the Probability of a hypothesis, h, and its denial, ~h, given the evidence, e, and everything else we know about the world, b, thus:

Craig’s first argument is (I think) to call out a notation error Mauro made:

You’ll notice that Bayes’ Theorem doesn’t require you to assign a probability to (h | e) as a means of determining Pr(h | e&b). Therefore it is plainly false that “in order to render the claim more probable than not, we need strong evidence in order to counterbalance the claim’s low prior probability (Pr (h/e) >>>> 0.5), which is just to say that we need ‘extraordinary evidence’ for such a claim.”

The only thing that is “false” here is the notation (in bold). The statement without that mathematical notation is true. So Craig is making a very confusing statement here. It looks like he is saying the maxim is false because there is no need to “assign” a value to Pr (h | e). But that cannot be what he means—as that would be mathematical gibberish, and a blatant non sequitur.

It’s hard to tell if this is a deliberate attempt to deceive, or just really phenomenally bad writing. But the only thing Craig could actually correctly mean here is: it is true that “in order to render the claim more probable than not, we need strong evidence in order to counterbalance the claim’s low prior probability, which is just to say that we need ‘extraordinary evidence’ for such a claim,” but it is incorrect to annotate that as Pr (h/e) >>>> 0.5; the correct notation is Pr (e/h) >>>> Pr (e/not-h). Mauro miswrote Pr (e/h) as Pr (h/e), and incorrectly wrote >>>> 0.5 instead of >>>> Pr (e/not-h). But instead of merely correcting the notation error and acknowledging the maxim is true, Craig appears to turn this into an argument against the maxim. That looks duplicitous.

Because Mauro did go on to correctly say that “strong evidence” can only be defined as evidence that is extraordinarily improbable if the extraordinary event had not occurred, which means (in correct notation) that Pr(e | not-h&b) must be extraordinarily low. That’s just not “the probability (h | e).” So that should have been the only point Craig made here. Otherwise, Craig’s first argument does not present any evidence the maxim he is claiming is false, is false. It’s merely a trivial statement about notation.

Defining “Extraordinary”

The only thing Craig says that could relevantly be an objection to the maxim “extraordinary claims requires extraordinary evidence” is in his following paragraph. Which begins by declaring:

What is needed to counterbalance any low prior probability of (h | b) is that e be much more probable on h than on not-h. You correctly note this factor in your penultimate paragraph.

This literally restates the maxim Craig declares false. So Craig is now affirming the maxim and thus contradicting himself. So what is going on here? Craig is playing a verbal game. Because he then immediately claims:

But then you erroneously infer, “But this. . . is equivalent to say[ing] that extraordinary claims require extraordinary evidence in order to overcome the prior improbability of the claim.” That is plainly wrong, unless one begs the question by simply defining “extraordinary” to mean “more probable on h than on not-h”

But that is not an inference from what Craig just affirmed. It is a literal restatement of what Craig just affirmed. So Mauro cannot have “erroneously inferred” his statement from Craig’s. Nor can Mauro’s statement just quoted be wrong. Much less “plainly” so. It’s actually right.

Craig seems to have only one actual argument then: that it “begs the question” to “define” the word “extraordinary” to mean “more probable on h than on not-h.” But that’s not begging the question. That’s defining the word. Mauro’s entire point is that this is what “extraordinary evidence” means: that “extraordinary claim” must mean a claim with an extraordinarily low prior; the only way to overcome that is with extraordinarily improbable evidence; therefore that is what we mean by “extraordinary evidence.” That is a correct inference. And it simply restates what Craig himself just affirmed is true.

Craig is thus not responding to Mauro’s point. He is instead lying: he is claiming that people who affirm this maxim don’t mean by “extraordinary evidence” evidence that’s extraordinarily improbable unless the claim is true. But Mauro just said that was what he means. And in actual fact it is what everyone means who uses this maxim. Craig cannot simply assert Mauro meant the opposite of what he said; nor can he claim this isn’t what everyone else is saying. As Mauro himself notes: “every single Bayesian analysis (besides yours) that I’ve read about [it] seems to support it.” Because all those analyses mean by “extraordinary evidence” exactly what Mauro means by it. Craig gives no evidence or reason to doubt this.

Nevertheless, Craig continues this bluff by claiming that, instead, “what the sceptic means by ‘extraordinary’ is something like ‘enormous or unusual’.” But that is just restating the same thing in different words. What does “unusual” mean? Improbable. They are literally synonymous. What does “enormous” mean? A lot. Why is “a lot” of evidence “extraordinary” evidence? Because having so much evidence is improbable, unless the claim is true. There is literally nothing else “sceptics” can mean by these words. And when we substitute for those words the only thing sceptics can mean by them, we get what Craig just admitted was true: “what is needed to counterbalance any low prior probability of (h | b) is that e be much more probable on h than on not-h.”

So Craig is not actually making any argument against the maxim here. He is instead lying, about what sceptics mean, claiming they mean instead something they cannot possibly mean, and then arguing that “that” meaning produces a false maxim. And we can be sure this is a lie, because there is literally nothing else sceptics could mean by “enormous or unusual” than “very improbable” (“unless h is true”). So there is no way Craig could honestly think they mean something else by it. Particularly as this is literally what Mauro just told him he means by it!

Craig’s “Argument”

So now that Craig has deliberately created a sham meaning for the maxim, something Craig dishonestly declares contrary to the meaning just told him by the person using the maxim, Craig goes on to beat down the straw man he dishonestly created:

In that sense the evidence need not be extraordinary. Otherwise we’d be forced to deny many non-miraculous events which we know to have occurred despite their high improbability. For example, I recently heard on the news that a particular person in South Carolina had won a powerball jackpot of over a billion dollars. I’m told that the odds of winning were something like one chance out of 320 million. So do I need extraordinary evidence to believe the report on the evening news that that particular person had won? Of course not! The evidence that it takes to counterbalance the low prior probability of that person’s winning needn’t be enormous or unusual at all but just more probable given the truth of the hypothesis than its falsehood.

Craig has pulled a rhetorical trick here (and someone has already pointed this out on his own website). He is claiming winning a lottery is an extraordinary claim, and yet we believe it on non-extraordinary evidence, “therefore the maxim is false.” And yet both those assertions are false. And thus so is his conclusion.

Winning lotteries is an extremely common, superbly-well documented phenomenon. It is therefore ordinary. It is not extraordinary. That is why there is a vast difference between saying a South Carolinan won a lottery, and saying a South Carolinan rose supernaturally from the dead. That people win lotteries has been supremely verified as a common occurrence. That people rise supernaturally from the dead has not. That’s what makes it an extraordinary claim. Not a “merely improbable” one.

Translated mathematically:

  • The prior probability of someone winning a lottery is low (say, millions to one against), but not extraordinarily low—as there are millions of people playing lotteries every day, so the prior probability of there being a lottery winner in any given month is not low, but as near to 100% as makes all odds. And we’ve reliably documented thousands of cases (and few ever turn out false). So many, in fact, as to make it one of the most ordinary claims around today.
  • But the prior probability of someone rising supernaturally from the dead is vastly lower than that—so low, in fact, that unlike lottery winners, we’ve reliably documented not even a single instance of it in thousands of years. Nor of any powers capable of it. Of course, setting aside the “one instance in thousands of years” that Craig wants to claim is well documented—but we mathematically must set it aside, as that cannot be included in the prior probability, which is conditional only on background knowledge (b), not the evidence of that case (e).

And when a claim is that improbable—having a prior so low it gives us an effective frequency of ~0 across all of human history (see How Not to Be a Doofus)—the evidence we would need to believe it happened has to be, as Craig himself admits, just as improbable (unless the claim is true). Hence, we need extraordinary evidence to believe extraordinary claims. Lottery wins are ordinary claims, so ordinary evidence suffices for them. But miracle claims are extraordinary claims. And that’s why they need extraordinary evidence. Which means, just as Mauro said, evidence that is extraordinarily improbable unless the claim is true.

How Extraordinary Evidence Actually Works

There are many different ways to get evidence that strong. You can, indeed, get there by having “a lot” of it—since the more evidence there is, the less likely it can all have arisen from other causes (like lies, errors, delusions, myths and legends, and the like), so eventually the amount of evidence can reach the scale of the extraordinary: which is when the probability that all that evidence would exist and the claim still be false is extraordinarily low. You can also get there by having a little evidence that’s extraordinarily “good.” Which means, evidence that even by itself is extraordinarily improbable unless the claim is true. Either is “extraordinary evidence.” And either or both would be sufficient to prove likely an “extraordinary claim.” Like rising from the dead.

Just continue the analogy of lottery winners. What if we lived in a world where magically rising from the dead were as common as winning lotteries? In that world, the amount of evidence proving resurrection a thing would indeed be extraordinary, and thus more than sufficient to establish that resurrection in that world is an ordinary claim. And still be as rare as lottery wins—just as with lottery wins. But nevertheless still as ordinary as lottery wins, and thus resurrections could be believed on just the same scale and quality of evidence as lottery wins are.

Much the same follows even if resurrections specifically were not so common or even yet known, but supernatural powers capable of effecting resurrections were. For instance, if we lived in Harry Potter’s world, like him, we would have acquired by now more than extraordinary evidence (in scale and quality) that magical powers exist. And so for us, that resurrection might be possible would no longer be an extraordinary claim. It might be extremely rare, perhaps unattested; but it would still be within the same realm of probability as teleporters are in our world now, and thus would require no more evidence to believe than we’d need to believe teleporters had been invented.

Craig has not only falsely claimed “sceptics” don’t mean by extraordinary “extraordinarily improbable,” and dishonestly misrepresented ordinary events (like winning lotteries) as just as extraordinary as miracles—thus dishonestly not refuting Mauro’s maxim while pretending to have done so—he also dishonestly misrepresents the evidence we have for lottery wins as not itself extraordinary! But alas, it is. If we had the same scale and quality of evidence for miracles as we have for lottery wins, miracles would be easily proved real.

And indeed, evidence for lottery winners is extraordinary. Think of the vast scale of evidence we have confirming that “winning a lottery” is a common phenomenon today, and the extraordinary improbability that all that evidence could be faked or mistaken. If we had that kind of evidence for the resurrection of Jesus—or indeed of any South Carolinan today—we would indeed be warranted in believing it. There is no sense in which the vast system of records and multiple verifications and checks against falsely reporting them, that warrant our trusting reports of lottery wins today, would not be extraordinary enough to verify at least a wondrous resurrection (even if we’d need yet more to confirm its cause).

And that’s not just true for “lottery winners” as a general phenomenon, but even for individual wins. Though for that the evidence we have might not qualify as extraordinary, but neither are claims of winning extraordinary, given all that vast evidence of the phenomenon in general. The evidence we have for individual lottery winners is still more improbable (unless those claims are true) than winning is; which is why we believe reported wins. So although individual lottery wins are not extraordinary claims, they are improbable, in the sense that few people do win; but the evidence available to us for each winner is more than improbable enough to satisfy the requirements of belief. Thus satisfying the maxim.

After all, what is the known base rate of media reports of lottery winners being false? It’s extremely low; much lower than the probability of any individual winning a lottery (indeed, can you even think of an example?). And how do we know the base rate of media-reported lottery wins being false is so low? A vast array of background knowledge about the social system we now reside in, where reported wins are commonplace, and it would be effectively impossible for the media to falsely report wins and not have been caught at it. (And the imagined conspiratorial systems required to make that even likely, are themselves extraordinarily improbable without any evidence of their operating, as all Cartesian Demons are.)

For how this all plays out when looking at the deeply problematic evidence sets we have for ancient times, see my talk on Miracles & Historical Method. And for further explication in respect to Craig’s dishonesty here, see my article Crank Bayesianism: William Lane Craig Edition.

Conclusion

Craig concludes:

There is no dispute here about Bayes’ Theorem. What is really at issue is the meaning of the term “extraordinary.” Sceptics seem to equivocate on its meaning in order to mislead people into thinking that an enormous or unusual amount of evidence is required in order to establish an event which has a high prior improbability.

It is not “sceptics” who are “equivocating” on the meaning of the term “extraordinary.” That is literally what Craig just did—despite knowing full well what Mauro actually meant by “extraordinary” (because Mauro literally just told him) and what Mauro understands all other “sceptics” to mean by it (because Mauro literally just told him).

Craig presents no evidence that any advocate of the maxim “extraordinary claims require extraordinary evidence” defines “extraordinary” in any way contrary to how Mauro just defined it right in front of Craig. And yet on Mauro’s own definition, Craig actually concedes that maxim is correct. But wanting instead to declare it false, Craig fabricates a nonsensical definition for “extraordinary,” dishonestly attributes it to Mauro and all other “sceptics,” and then makes a completely fallacious argument even from that fabricated premise. And there is no way this can be an honest mistake.

That’s William Lane Craig’s entire life schtick in a nutshell.

The bottom line is this:

Craig admits “what is needed to counterbalance any low prior probability of (h | b) is that e be much more probable on h than on not-h” and therefore Craig admits the maxim “extraordinary claims require extraordinary evidence”—as actually defined by advocates of that maxim—is true. Not false. And his attempt to nevertheless claim it is false is entirely built out of dishonestly false definitions, and blatant errors in determining the probability of ordinary events and evidence.

Winning lotteries is not extraordinary. It is in fact so incredibly common and widely documented as to be totally ordinary. Moreover, we actually do have extraordinary amounts of very strong evidence that news reports of winners are almost always true. For miracles, neither is ever the case. They have never been verified as commonplace factual events the way lottery wins have; and we have no evidence for any miracle even remotely as vast or strong as the evidence we have for the reliability of reported lottery wins.

And that is why the maxim “extraordinary claims require extraordinary evidence” is demonstrably true, and not false as Craig dishonestly attempts to claim; and why it effectively refutes all miracle claims to date. All of which are genuinely extraordinary; yet for none of which is any evidence so.

§

To comment use Add Comment field at bottom or click a Reply box next to a comment. See Comments & Moderation Policy.

Discover more from Richard Carrier Blogs

Subscribe now to keep reading and get access to the full archive.

Continue reading