Monday, November 06, 2006

Cooking the books

Bill Curry* has posted a sequel to his article on “Evaluation the Evidence for the Resurrection.”

http://debunkingchristianity.blogspot.com/2006/11/evaluating-evidence-for-resurrection.html

*For now I’ll accommodate this conventional pseudonym. See below for my source-critical analysis, in which I peel away the redactional layers to uncover the Ur-Curry.

I’ll content myself with a few comments.

Curry says:

“I fleshed this out in greater detail here and argued that the background probability that God would raise Jesus from the dead (compared to the legendary hypothesis) should be no higher that 1 in million.”

Except that if you go back to his earlier article, he doesn’t presented a detailed argument for this figure. Indeed, he doesn’t present any argument whatsoever. This is what he actually says:

“It certainly doesn’t seem unreasonable for a historian to think (in the absence of evidence) the probability a reported event requires supernatural cause is one in a million 10-6.”

But that’s an assertion, not an argument—much less a detailed argument. There’s a complete failure to explain how he arrived at that figure. We’re simply told that it “doesn’t seem unreasonable.”

This is one of the problems with his appeal to Bayesean probability theory. Even if we accepted that framework, Curry simply mouths a certain amount of Bayesean verbiage to create the illusory impression of quantitative rigor.

But that’s all we get. Some Bayesean lingo. Some fractions and percentiles. But where’s the supporting argument? Where’s the actual process for generating these figures?

So the Bayesean appeal is just an empty show. He’s trying to wow the reader by dropping some technical language into his article, and sticking some numbers here and there, but the actual computations are missing.

Like sawing the lady in half at the circus, the Bayesean trap door is kept out of public view. All we get to see is what the magician wants us to see. It would spoil the illusion to know what’s really going on.

Continuing:

“So far I have considered 1) the time between the reports and events and 2) chief priest's need of Judas as relevant evidence with regard to determining if the Gospel accounts are legendary or historical (see here).”

I’ll confine myself to (1). And if we go there, what do we find?

***QUOTE***

Richard Carrier relates the story of Saint Genevieve here

In 520 A.D. an anonymous monk recorded the life of Saint Genevieve, who had died only ten years before that. In his account of her life, he describes how, when she ordered a cursed tree cut down, monsters sprang from it and breathed a fatal stench on many men for two hours; while she was sailing, eleven ships capsized, but at her prayers they were righted again spontaneously; she cast out demons, calmed storms, miraculously created water and oil from nothing before astonished crowds, healed the blind and lame, and several people who stole things from her actually went blind instead. No one wrote anything to contradict or challenge these claims, and they were written very near the time the events supposedly happened--by a religious man whom we suppose regarded lying to be a sin.

***END-QUOTE***

But the comparison with the gospels breaks down at several key points:

1.The life of St. Genevieve as anonymous, whereas, as Martin Hengel has argued (see the quotations by Jason Engwer), the Gospels are not. By knowing who wrote the gospels, we’re in a position to know what they were in a position to know.

2.The life of St. Genevieve lacks multiple-attestation, unlike the NT record of Christ.

3. Some of the miracles are clearly imitative of dominical miracles. So this is a secondary, hagiographic embellishment.

Curry cites two more cases:

“There are more modern examples. Many still believe that the crash in Roswell, New Mexico in 1947 is evidence of UFO's. This is in an era when the government keeps records and people have access to the Internet.”

Notice the equivocation here. Who believes in a gov’t cover-up at Roswell? People who were actually on the scene in 1947? Eyewitnesses? People who interviewed eyewitnesses?

Or are we talking about modern-day gossip-mongers at several removes from the original event, who rely on third-hand information?

One more case:

“The fact that Benny Hinn has made miracle claims and still has a following persuades me that even today, people are willing to accept legends as true in a very short time.”

Now why do you think that Curry brings up Benny Hinn? He assumes that the reader will dismiss Benny Hinn as a charlatan.

And why would he make that assumption? Because Benny Hinn’s supernatural claims are widely disputed. Disputed to the point where Curry can depend on the reader to dismiss Benny Hinn as a fraud.

How is that comparable to the Gospels? The charge leveled by hostile sources like Celsus and the Talmud is not that Christ was a fraud, but a sorcerer. So his miraculous powers are not denied, but attributed to the dark side.

Going back to Curry’s latest article:

“I present numeric values in these assessments primarily as a tool to help me think in a disciplined way, not to provide a precision to the results.”

But if they don’t provide precision analysis, why dabble with the Bayesean apparatus?

Observe some of his claims:

“I would assess that P(E|Resurrection)/P(E| Legend) ~ 1.6 (about +2 dB).”

“I would assign this ratio the value of 0.5 (about -3 dB).”

“Reported conversion of James. Assessment posterior: 0.96 x 10-6”

“Jesus' ignorance deemphasized. Assessment posterior: 0.48 x 10-6”

So he presents the reader with results on the order of such-and-such fraction to the negative power of such-and-such.

Isn’t this a transparent attempt to create the impression of mathematical precision?

But then he prefaces his results with this general disclaimer.

And I can understand why. For, once again, no effort is made to justify the assignment of these “plausibility” ratios and percentiles by actual computation.

So what are we to make of these figures? Are they just deceptive window-dressing? Cosmetic figures to foster a false impression of mathematical rigor when, in fact, all the hard work is left out of the equation?

The reason he avoids the detailed questions is that it would immediately expose all of the unspoken and unsupported assumptions which are feeding into his ersatz results. As one philosopher explains:

***QUOTE***

Indeed, any Bayesian analysis of the question of justified belief in miracles must be otiose until the difficult and essential questions concerning "evidence" in relation to an allegedly miraculous occurrence are resolved — at which point any Bayesian analysis will add little except the technical complexity of a formal apparatus that may or may not "clarify" the structure of Hume's argument.

The balancing of probabilities is of no use until it is decided what goes into the balance — that is, what constitutes the evidence that is to be subject to the balancing of probabilities. The point is this; apart from independent philosophical arguments — arguments that would in effect undermine the relevance of a Bayesian analysis to the question of the credibility of reports of the miraculous — no such analysis can, in principle, prove that no testimony can (or cannot) establish the credibility of a miracle.

http://plato.stanford.edu/entries/miracles/

***END-QUOTE***

But two can play this game. Permit my to my own back-of-the-envelop calculation. Assuming that there were about 6 billion people alive today, there’s only a one-in-six billion chance that Bill Curry even exists.

Given such overwhelming odds against his existence, I don’t believe that Curry is a real person. Just imagine the extraordinary evidence it would take to overcome such a crushing presumption to the contrary?

It is far more rational to subscribe to the Curry-myth. According to the mythic hypothesis, Curry never existed.

What is more, there are certain anomalies in the Curry-myth which point to legendary embellishment—such as the pretense of being a Bayesean probability theorist when, in fact, there’s no hard evidence that the pseudo-Curry is more than a dilettante in this highly specialized field of probability theory.

Hence, it’s far more rational to conclude that “Bill Curry” is a literary construct. John Loftus needed to dress up his infidelity in a pseudoscientific garb of mathematical precision, so he invented the character of Bill Curry, as well as his brother Jon.

But the manifest lack of technical competence in the actual elaboration of the argument betrays his educational limitations.

This is further corroborated by the fact that Jon and Bill are stock characters, recycling the stereotypical anti-hero in the monomythic saga of apostasy. Their utterly predictable, cliché-ridden, and one-dimensional characterization is just what you’d expect of a literary construct by a hack writer like Loftus. Pulp fiction in the service of atheology.

Finally, we have this delicious statement:

“The author of Matthew also covers roughly the same event. Here the phrase "nor the Son" is missing in Matthew 24:36 (or it is at least not present in early manuscripts according to my NIV). Matthew's omission is consistent with legendary development and less consistent with the claim that Matthew was writing history.”

Notice how he slides over the distinction between an original, editorial omission and a subsequent, scribal omission. But a scribal omission would not be evidence that Matthew was writing something other than history.

In fact, Bart Ehrman, for one, who’s no friend of inerrancy or Evangelicalism does, indeed, attribute the omission to scribal tampering. Cf. The Orthodox Corruption of Scripture (Oxford 1993), 91-92.

And while we’re on the subject of omissions, Curry doesn’t bother to interact with Stephen T. Davis’ criticisms of Bayesean theory in this context:

http://www.philoonline.org/library/davis_2_1.htm

And it won’t do just to refer us to Martin’s rejoinder, for we must still evaluate the degree to which, if any, Martin successfully defangs his critic.

But I look forward to Loftus’ next installment in the apocryphal Acts of the Brothers Curry.

6 comments:

  1. Steve,

    On the general subject of probability, I wanted to point out what I think is an error in your ebook “This Joyful Eastertide”.

    You say:

    **********************
    “The odds of a royal flush are about 1 in 650,000 whereas the odds of a straight flush are about 1 in 72,000. This means the odds of drawing 9 straight flushes in a row are about the same as drawing one royal flush. But while I could get away with a royal flush, were I to draw 9 straight flushes in a row, casino security would be fitting me with a pair of concrete galoshes.

    Probability theory is unable to capture certain common sense intuitions, especially where personal agents are in view.”
    **********************

    I actually think that our intuitions are quite consistent with the mathematical facts in this case. The probability of 9 straight flushes in a row is NOT 1 in 72,000 x 6 = 648,000. It is, rather, 1 in a whopping 72,000 to the 9th power!

    In the history of poker a royal flush has been drawn on many occasions, I'm sure, but I’d wager my “nest egg” that an honest 9 straight flushes in a row has never been drawn and never will be.

    Turns out those thugs with the concrete galoshes know their math.

    Just a side note.

    ReplyDelete
  2. Readers might be interested in knowing that I interacted at length with Bill Curry's first article in his current series, and that he eventually changed his argument about Judas, then left the discussion. Here's the URL for my first response to him:

    http://triablogue.blogspot.com/2006/09/error-in-error-out.html

    Then he responded with another article:

    http://debunkingchristianity.blogspot.com/2006/09/jason-engwer-responds-to-evaluating.html

    And here's the final thread, in which he left the discussion after changing his argument about Judas:

    http://triablogue.blogspot.com/2006/09/gospels-as-historical-accounts.html

    Notice that he not only changed his argument about Judas, but also was shown to be wrong and relying on bad sources on other issues. People might be interested in reading what I wrote there about Saint Genevieve. And I linked to an article I've written that addresses the comparison between Jesus Christ and Benny Hinn:

    http://triablogue.blogspot.com/2006/05/jesus-christ-benny-hinn-and-santa.html

    ReplyDelete
  3. Not to be ignorant, but what are 'dominical' miracles?

    ReplyDelete
  4. "Dominical" is an adjective which has reference to things said or done by the Lord (Dominus) Jesus Christ.

    ReplyDelete
  5. Steve’s Quote: . Permit my to my own back-of-the-envelop calculation. Assuming that there were about 6 billion people alive today, there’s only a one-in-six billion chance that Bill Curry even exists.

    Steve, I think this comment indicates a fundamental misunderstanding of Bayesian inference. But here is a chance for you to prove me wrong. Let’s use a more numerically extreme example. Let’s suppose we toss a coin 1000 times. The probability of any observed sequence is 2 raised to the -1000 (about 9.33e-302) if the coin is fair. Why would a Bayesian conclude the coin was fair given that the observed sequence is so improbable? Do you think this is paradoxical on the Bayesian view? I obviously don’t think this causes any problems for the Bayesian. Do you know why I think this?

    Please answer these questions. In my opinion, you subject your opponents to a lot of ridicule, but the ridicule may be worth it if I am going to learn something from you. At this point, I am unconvinced that you are making a reasonable effort to understand my position or Bayesian inference in general. Your behavior is very surprising if one thinks that the Holy Spirit is working in your life. I certainly hope I and my own children can refrain from ridiculing those they disagree with as you regularly do here.

    I comment on this quote when you first posted it. Indeed, any Bayesian analysis of the question of justified belief in miracles must be otiose until the difficult and essential questions concerning "evidence" in relation to an allegedly miraculous occurrence are resolved — at which point any Bayesian analysis will add little except the technical complexity of a formal apparatus that may or may not "clarify" the structure of Hume's argument.

    The balancing of probabilities is of no use until it is decided what goes into the balance — that is, what constitutes the evidence that is to be subject to the balancing of probabilities. The point is this; apart from independent philosophical arguments — arguments that would in effect undermine the relevance of a Bayesian analysis to the question of the credibility of reports of the miraculous — no such analysis can, in principle, prove that no testimony can (or cannot) establish the credibility of a miracle.


    It seems to me that you are mischaracterizing what the author is saying here. The author is indicating that Bayesian analysis is not applicable for assessing the question of miracles in general (Hume’s argument in particular), with which I agree. In principle Bayes’ Theorem could be used to support particular miracle claims. But the author did not make the case that Bayesian inference cannot be used to assess a particular miracle claim. The author may or may not believe that, but this quote doesn’t strongly support the idea that Bayesian inference is not a useful tool for such assessments.

    But two can play this game.
    If you are interested in a valid assessment I hope that you would. I have attempted to make my argument as scrutable as possible. I would welcome you doing the same. If you would like, I could email you or Jason the MS Excel spreadsheet I used so you could use that as a template for your inferential argument for the resurrection. Of course, if you want to obfuscate the issues as much as possible, this would not help you achieve your goals.

    Curry doesn’t bother to interact with Stephen T. Davis’ criticisms of Bayesean theory in this context.
    Because MacKay and Jaynes (both books published in 2003) have successfully defended Bayesian inference to my satisfaction. Some of the criticism that you have offered have been variants of the Hempel paradox which has been answered as early as 1967 by I. J. Good, according to Jaynes.

    I disagree with some of the number’s Martin used as well, but I would agree with Davis that “I have said little thus far about the way Martin uses Bayes's Theorem, which by and large is beyond reproach, at least until he starts supplying actual values.”

    I think my a priori assessment for the initial plausibility is quite reasonable. Pagan miracle claims vastly outnumber miracle claims in the gospels and far exceed the number of miracles observed today. This fact in conjunction with the fact that there is a great deal of known fraud in religious writing gives me to think that my a priori expectation of miracles is quite reasonable. To quote MacKay “you can’t do inference – or data compression – without making assumptions.” I have tried to make my assumptions clear, will you?

    Jason,

    I don’t think I changed my argument, and I don’t think you interacted with what I said. I know you disagree. However, if you want to move the conversation forward, why don’t you present your own Bayesian argument for the resurrection? If you want, I can provide you the spreadsheets I used. I really get the impression that you are obfuscating here. But if you are trying to clarify the issues, show us how the Bayesian inference should be done. What is the evidence that should move truth seekers, and how much should that evidence move us? The spreadsheets help my though processes. They may not help everyone, but if you want to reach those mathematically inclined, I would encourage you to lay-out your assumptions and assign plausibility values.

    ReplyDelete
  6. Bill Curry said:

    "I don’t think I changed my argument, and I don’t think you interacted with what I said."

    Your original argument was that the religious leaders wouldn't have needed Judas to identify Jesus for them. After that argument was shown to be bad, you shifted your focus to the argument that Judas could have been setting up an ambush, so that it's unlikely that the religious leaders would have worked with him as the gospels describe. I explained why both arguments were unreasonable. You never responded to my last post on the subject. You also failed to interact with the evidence I cited on some other issues, such as whether the gospels were anonymous.

    I'm not going to make calculations with Bayes' Theorem. It's unnecessary, and there are better ways to use my time.

    ReplyDelete