Monday, July 22, 2013

Proving History?

So I finished reading Richard Carrier's book "Proving History: Bayes's Theorem and the Quest for the Historical Jesus" which I mentioned in the previous post.

A few things to say about Richard Carrier before I make comments on this book in particular:
  1. Carrier is an intelligent guy who thinks things through in greater detail than many people do.
  2. He also comes across as quite arrogant in interviews and in some of the things he writes.
  3. His logic appears to be sound and his conclusions generally appear to be justifiable.
  4. He has a clear idea of who his audience is, and tailors his writings to that specific audience.
Starting with that final point, I don't think I am the audience Carrier is writing to in this book. In many of the books by 'new atheists' which I have read, I've realised that the authors are generally 'preaching to the deconverted', as it were, and are not realistically expecting many believers to be reading their books. 'New atheism' books are aimed to give ammunition to atheists. Of course, the flipside of that is that most apologetics books are aimed at Christians to reinforce their beliefs, and not to try to convince atheists to believe...

So, point number one in this review, I don't think any Christian is going to be swayed by the arguments made in this book, even if they (the arguments, that is) are logical, justified and, indeed, correct. But, of course, this book doesn't really set out to sway anyone with its conclusions about history, the main point of this book is to demonstrate to its (mostly non-mathematically inclined) audience that the mathematical technique of Bayes's Theorem is appropriate in historical investigation, and furthermore is the best method to use in historical investigation. I think the book manages to do that, but as conclusions go, its not a very exciting one. The much more exciting and dramatic conclusion comes in Carrier's next book, which isn't out yet.

As a side note, I'm glad that Carrier settled on consistent use of "Bayes's Theorem" rather than "Bayes Theorem", which I find acceptable, or "Bayes' Theorem", which is clearly not acceptable to anyone who understands grammar. Of course, he abbreviates it to "BT" for most of the book, which gets around the problem of the apostrophe. But anyway...

On the whole, this is not a very interesting book. It has interesting bits in it, and makes some interesting conclusions along the way, but I can't really recommend this to anyone as a good read. I doubt even Carrier thinks this book is particularly interesting, but it probably is required before we get to the meaty stuff in the next book. And therein lies the problem, this is clearly the build up to the much more interesting work which is to follow, and may be required reading to understand that second work, but by itself this is a bit dry.

I guess I should explain what Bayes's Theorem is. Put simply, it is a mathematical method for updating your beliefs in the light of new evidence. As used by Carrier, it is a method of assessing the probability of one or more historical hypotheses by considering one or more pieces of evidence. The method provides a framework for determining whether our confidence in a given belief (hypothesis) should be increased or decreased as new evidence presents itself. (I should note here that I have used Bayes's Theorem in my work for over a decade now, so I am fairly familiar with it, and thus am far from the target audience of this book.)

The point of this book is to convince the reader (who Carrier expects to be a non-mathematically inclined person with an interest in history) that Bayes's Theorem is a sound and justifiable method for coming to historical conclusions. He wants everyone to share this belief. If he gets everyone to share this belief, then everyone will have to accept any historical deduction he makes using Bayes's Theorem. And in the next book (I expect) he will end up using Bayes's Theorem to show that Jesus was not the Son of God and, furthermore, probably didn't exist. But in order for that argument to work, he has to get you to accept the method.

The fundamental problem in this kind of reasoning is this, that faced with an anti-Christian conclusion derived using Bayes's Theorem, the non-mathematically inclined Christian historian who accepted BT on the basis of this book will reject it again after the next book, because their acceptance of the truth of Christianity is so much stronger than their acceptance of the truth of Bayes, that the only conclusion they can come to is that Bayes's Theorem must be flawed, even if in some way that is not immediately apparent. So I fear that the end result of this will be an insufferably smug Richard Carrier, who knows - beyond reasonable doubt - that he has proved that Jesus didn't exist, and a bunch of Christians who have read this, who know that Carrier must be wrong.

But Carrier does an excellent job of demonstrating that Bayes's Theorem is a valid way of estimating historical probability, and a pretty good job of demonstrating that it is the best method of estimating historical probability.

Perhaps the most interesting chapter in the book is Chapter 5, where he looks at the various criteria used by historians involved in the 'quest' for the 'Historical Jesus' and either demonstrates that each criterion reduces to Bayes's Theorem, or is invalid as a method. Thus the point is proved, Bayes's Theorem is the fundamental valid method underlying all other historicity criteria.

Taken one by one:
  1. Dissimilarity: demonstrated to be invalid.
  2. Embarrassment: similarly invalid. Carrier deconstructs this one at great length, indeed, it is the largest part of the chapter by far. Along the way, he makes a very interesting statement that I must investigate further sometime:
    "Its worth remarking here, ..., that everyone literate enough to compose books in antiquity was educated almost exclusively in the specific skill of persuasion: that is what all writing was believed to be for, and how all literate persons were taught to write." (page 134)
    Thus, everything in an ancient book was there for the purpose of convincing the reader of something, so nothing embarrassing which hindered this purpose would be included.
  3. Coherence: shown to be invalid.
  4. Multiple attestation: shown to be invalid.
  5. Explanatory credibility: consistent with Bayes, but can only exclude, never confirm.
  6. Contextual plausibility: similarly consistent with Bayes.
  7. Historical plausibility: again, consistent with Bayes, but is incomplete.
  8. Natural probability: consistent with Bayes, but doesn't make as strong a case as could be formulated with Bayes, so Bayes supersedes this method. 
  9. Oral preservability: consistent with Bayes, but can only exclude.
  10. Crucifixion (the theory has to explain why Jesus was deemed worthy of crucifixion): not valid.
  11. Fabricatory trend: consistent with Bayes, but can only exclude.
  12. Least distinctiveness: of limited use, but consistent with Bayes.
  13. Vividness of narration: shown to be invalid.
  14. Textual variance: invalid.
  15. Greek context: invalid.
  16. Aramaic context: invalid.
  17. Discourse features: invalid.
  18. Characteristic Jesus: invalid, as it relies on many of the above invalid methods.
Wow. I hadn't realised there were so many rubbish criteria in use by Historical Jesus scholars. Carrier shows that most of them are useless and the rest of them are poor-man's versions of Bayes, so to use Bayes directly is better than using any of them.

Anyway, I think I've said all I need to here. There's an interesting Jesus vs Daniel thing that I'll maybe pick up in a future post, but I'll not go into it here.

In summary, I agree with Carrier's conclusion: "Historians should be Bayesians". We'll see what happens when Bayes is applied directly to Jesus in the next book by Carrier.


Monday, July 01, 2013

Historical probability and the Son of God

I'm currently reading Richard Carrier's book "Proving History: Bayes's Theorem and the Quest for the Historical Jesus". A proper review may follow in due course, when I've finished reading the thing. Carrier is currently working on another book in which he aims to prove that Jesus never existed at all, using the methods he establishes and validates in the earlier book. But we'll have to judge that book on its own merits, when it comes out.

Here, as the title makes clear, Carrier uses Bayes's Theorem as his primary method for assessing historical hypotheses in the light of historical data. Now, I'm no stranger to Bayes's Theorem, it was the fundamental foundation of my PhD thesis (in Engineering) after all, and I'm very interested in the subject of Jesus, historical or otherwise, so I really had to read this book. 

Currently I'm about 1/3 of the way through the book, so I can't comment on where Carrier ends up, but I have a nagging doubt about the applicability of the use of Bayes's Theorem, or indeed any method of estimating historical probability here.

The problem is this. It doesn't matter which way you slice the issues, or which pieces of evidence you consider or neglect, you will never, ever reach the conclusion that Jesus was the Son of God by historical methods.

Almost by definition, Jesus, if he was the unique Son of God (whatever that means in this context), was just that, the unique Son of God. There was only one of him who ever lived. That means that historically there is only a 1 in 100 billion (approx number of people who have ever lived) chance that any given person in history was the real Son of God. Now those odds are so vanishingly small that any probabilicist would quite happily neglect that chance and simply say the chances of it are zero.

In other words, historically speaking, the chances of anyone being the Son of God are zero. Therefore, in terms of historical probability, we can say with a great degree of confidence that nobody was the Son of God, therefore, Jesus was not the Son of God. QED. Probability does work like that. But the question is whether or not reality works like that? 

My doubt is that if Jesus actually was the 1 in 1011 Son of God, then historical reasoning will never give us access to that fact. We simply cannot ever get there through historical method.

So when it comes to Carrier proving mathematically that Jesus was not the Son of God and, furthermore, did not exist at all, then that really proves nothing. The issue will be completely settled for a number of atheists who believe it anyway, but it will change nothing for even mathematically minded and rational believers (yes, they do exist), because there is always a difference between a 'vanishingly small' probability and actual zero. And it doesn't matter how small 'vanishing' is, Bayes's Theorem will always leave a tiny hole for that chink of light to shine through, for those who believe in the light.