Very nice overview of Bayes Theorem and Historical Jesus

[_Analytics]

Assuming that there is something to the historical Jesus, which existed first, the oral stories of Jesus told by eyewitnesses or the Pauline epistles? Were the gospels written based on myth or the fact that the eyewitnesses were dying off and the early Christians wanted to preserve their stories.

If there was a historical Jesus, then of course the oral stories began as the historical life of Jesus began to unfold. "Wow! Did you hear what he just said! Wow, he just healed that guy!" etc.

But its odd that nobody that we know of would have thought to write anything down until decades later when the actual witnesses were dying off.

[_Analytics]

There's nothing wrong with Bayesian reasoning in scholarship like history per se. When I see it misused, there's usually a couple of problems at work in my experience.

First, people understate the vast base of background knowledge that forms our judgments, which is ironic given that this Bayesian reasoning here, and as a result, overvalue pieces of evidence they are talking about. Bayesian references end up being a fig leaf to obscure a shallow consideration of the total argument space. Ever read an evangelical apologist express the "Lord, Liar, or Lunatic" argument in Bayesian reasoning?

Second, people do a really poor job estimating probabilities at the margins and as a result sneak in highly illicit probability judgements that can affect downstream calculations. For example, when we find something intuitively unlikely, it's easy to accept a .001% probability instead of .000000000001% that it actually is. It's hard to put a number on a lot of things. We just don't make those fine distinctions well. That can matter in a complex enough stack.

Having read Carrier attempt a ridiculous Bayesian argument on a different subject not that long ago, he was guilty as sin when it came to the first problem. I'm not sure if it infects his other stuff, but I'd be willing to bet it does.

Carrier did write a 300 page book explaining his methodology in detail, and followed it up with a 700 page book that goes over all of the evidence he considers. These were attempts at serious academic writing, and shouldn't be dismissed simply because of what he writes on a blog.

He does spend quite a bit of time explaining that the more specific you make your hypothesis, the necessarily lower the a priori probability of it being right actually is. He spends quite a bit of effort in making the historical Jesus hypothesis as broad as possible. He does likewise with the mysticist hypothesis.

He gives wide ranges to the probabilities he assigns to various things, and at the end he says the probability of a historical Jesus could be as high as 33%.

On the one hand, you could argue in an over-specific manner a la "Lord, Liar, Lunatic." On the other extreme, you could be over-general and start sounding like Hugh Nibley claiming that "the ancients" were similar to Mormons in almost all respects. If Carrier is shoveling crap, it's going to be more like Nibley. Carrier creates very broad definitions of words like "crucifixion", and paints broad pictures of different groups and beliefs to make them look as similar as possible to what Paul said.

[_Philo Sofee]

And he also explains that his definitions of some words he uses such as crucifixion are exactly the way the Ancients themselves were using those words. To me that seems to be historically responsible scholarship.

[_Analytics]

...I would definitely side with Reverend Kishkumen, but I'd take it farther, and say that Bayes Theorem is totally useless for any field for the same reason Ockham's razor is. For the amusement of Analytics, I'll point out that my philosophy of science is rooted directly in the Efficient Market Hypothesis (that's from economics, for others). In that view, Bayesian calculations and appeals to simplicity find an analogy to metrics such as a Price/Earning ratio. Suppose the price of a stock is X and the P/E ratio is Y, and according to P/E, X is way to high. In one view of things, the market must be wrong since the price is not justified by solid company fundamentals, but in the EMH view, all the influential market players know what a P/E is, and so the stock price must be driven by something the market knows but that the metrics aren't capturing....

Amusement noted.

If there were a market trading on the likelihood of the historical Jesus hypothesis, and if the market odds said there was a 50-50 chance it was true, I'd probably take the long position and say sure, he probably existed. But practically everybody loves the historical Jesus hypothesis, so if everybody had a vote, the market price might be 99-1. In that case, I'd take the contrarian view. Did he exist? Maybe. Are the people who think he existed overconfident? Absolutely. That's where my money would be.

The EMH reminds me of an experience I had at Michigan State back in the day. I was walking across campus discussing efficient markets with my professor, and I noticed that there was a $20 bill lying on the grass underneath a maple tree. I pointed to the bill told him to pick it up since he was closer. Without even looking down he said, "Impossible. If there was a $20 on the ground, somebody would have picked it up by now."

[_Analytics]

And he also explains that his definitions of some words he uses such as crucifixion are exactly the way the Ancients themselves were using those words. To me that seems to be historically responsible scholarship.

I get that point and don't disagree with it, but I always get nervous when people's arguments depend upon generalities about "the ancients."

[_Symmachus]

Interesting thread. I have to say that I don't really see the point of this Bayesian theorem to historical work for a couple of reasons. One is that it doesn't seem to introduce anything that one can't already learn from traditional methods. Or, to be slightly more accurate, it doesn't seem to introduce any historical question that one can't ask without this method.

Added to that, it seems like a grand exercise in confirmation bias (at least from what I understand as I've seen it described here), since before one can apply it one must work within traditional terms in order to establish the question. That means, in other words, that one already has basically to choose a way to read the ancient evidence (as Kish has so thoroughly shown), a process which actually requires committing to a certain position. But that position, of course, is what the Bayesian theorem is supposed to help settle. So isn't there something circular to this?

Also, those of you who are adepts, is this supposed to have any predictive power? Using probability theorems retrojectively seems kind of ridiculous—or, at least, drawing any meaningful conclusions from it surely is ridiculous. Answering a question like "What is the probability that X was the case" really can't help you know whether X was in fact the case or not. Highly improbable things have happened (e.g. Sarah Palin's VP candidacy). But has anyone ever used Bayesian probability to predict some historical fact for which there was little to no evidence and had it confirmed or corroborated by evidence discovered later? And is there any other historical question to which this kind of reasoning is or has been usefully applied? Or is it just historical Jesus?

[_EAllusion]

Carrier did write a 300 page book explaining his methodology in detail, and followed it up with a 700 page book that goes over all of the evidence he considers. These were attempts at serious academic writing, and shouldn't be dismissed simply because of what he writes on a blog.

I don't think one should ignore Carrier because he uses terrible, though confidently asserted Bayesian reasoning in his blog. I do think it is a minor red-flag to consider given that his other work relies on Bayesian arguments. That Carrier is viewed as a fringe academic and seen about as seriously as global warming skeptics are viewed by climatologists by respectable scholars is a good reason to dismiss him if you want to. There's only so much time in the day. If you don't want to and are interested in analyzing his arguments that's fine too.

I'm not sure how much Bayes helps here. Did you find Bayesian calculations adding any insight into his historical arguments by generating surprising, counter-intuitive results or was it just an additional step that at best better articulates his ordinary historical arguments and at worst obfuscates them?

[_Analytics]

...Added to that, it seems like a grand exercise in confirmation bias (at least from what I understand as I've seen it described here), since before one can apply it one must work within traditional terms in order to establish the question. That means, in other words, that one already has basically to choose a way to read the ancient evidence....

No. The Bayes approach is to look at all of the evidence, and then evaluate how consistent it is with competing hypotheses. That's what a competent analysis would do anyway--expressing it in Bayesian terms just formalizes the reasoning.

The heart of Carrier's argument is essentially what Earl Doherty argues in The Jesus Puzzle. Doherty makes the same basic arguments without couching it in Bayesian terms.

Also, those of you who are adepts, is this supposed to have any predictive power? Using probability theorems retrojectively seems kind of ridiculous—or, at least, drawing any meaningful conclusions from it surely is ridiculous....

Bayesian analysis is a logically valid way of evaluating uncertainty. It's not a magic bullet, and garbage-in-garbage-out applies.

As an example, the web actually has a calculator called, "The Famous Bayesian-Moroni Prayer Analysis Calculator."

http://www.lds4u.com/lesson1/bayesian.htm

If you click behind all of the ads and look at the content there, you should see that this is a better approach for interpreting your emotional/spiritual responses to the Book of Mormon than simply assuming that any vaguely good feeling is God's way of telling you the Book of Mormon is true.

[_EAllusion]

Bayesian analysis is a logically valid way of evaluating uncertainty. It's not a magic bullet, and garbage-in-garbage-out applies.

As an example, the web actually has a calculator called, "The Famous Bayesian-Moroni Prayer Analysis Calculator."

http://www.lds4u.com/lesson1/bayesian.htm

If you click behind all of the ads and look at the content there, you should see that this is a better approach for interpreting your emotional/spiritual responses to the Book of Mormon than simply assuming that any vaguely good feeling is God's way of telling you the Book of Mormon is true.

The crazy luantic who created that website needs to update his web design.

Having seen this a few times before from mysterious sources, I must say that I bet you anything people culturally influenced by Mormons as a group drastically overestimate their sense of the prior probability the Book of Mormon is "true," even if they pick a low starting number. I worry that people may come away thinking teh maths proves their poor intuitions right.

[_Analytics]

Carrier did write a 300 page book explaining his methodology in detail, and followed it up with a 700 page book that goes over all of the evidence he considers. These were attempts at serious academic writing, and shouldn't be dismissed simply because of what he writes on a blog.

I don't think one should ignore Carrier because he uses terrible, though confidently asserted Bayesian reasoning in his blog. I do think it is a minor red-flag to consider given that his other work relies on Bayesian arguments. That Carrier is viewed as a fringe academic and seen about as seriously as global warming skeptics are viewed by climatologists by respectable scholars is a good reason to dismiss him if you want to. There's only so much time in the day. If you don't want to and are interested in analyzing his arguments that's fine too.

I'm not sure how much Bayes helps here. Did you find Bayesian calculations adding any insight into his historical arguments by generating surprising, counter-intuitive results or was it just an additional step that at best better articulates his ordinary historical arguments and at worst obfuscates them?

I don't think the Bayesian calculations add any insights in-and-of themselves. However, it is the correct logical structure for evaluating the evidence and making decisions in the face of uncertainty.

Does Bayes' help here? In this thread at least, it's done more harm than good, because it's distracted from Carrier's (and Doherty's) actual arguments.

Part of my bias here is that 20 years ago I registered for an Econometrics course in graduate school, and the professor assigned a textbook called "Bayesian Econometrics." Previously, I had thought of Bayes' theorem as a tangential topic towards the back of any Statistics textbook. Then this book was sprung on me that claimed that it was much more than that, and was a more "unified and coherent way of thinking about decision problems and how to solve them using data and other information."

These seemed radical at the time, but it seems to be slowly catching on--there are certainly more books about "Bayesian Econometrics" now than there were then.