Is an Appeal to Consensus an Argument?

Is an Appeal to Consensus an Argument? May 17, 2019

Christoph goes on to tackle the main theme of this post, i.e. the attempts of some to contrast expert consensus with crowdsourcing on the internet, characterized as the “democratization of knowledge.” Christoph continues in response (with the relevant quotes given):

“Welcome to the democratization of information and knowledge sharing, along with the opportunity to contribute specialist skills and insights.”

I appreciate that sentiment. (I really do!) I just found that for myself it is too time-consuming to get into these kinds of things and I try to find other opportunities for advancing the same cause. So please accept my apologies that I can’t respond in as much detail to your comment as would actually be necessary. That I spoke of “translation” is also such a case of too less time for this kind of comment. This was not the best way to put it. Let me try to summarise my position as succinctly as I can: you are right, I did not offer evidence on syntax and lexical semantics. In fact, it was my aim not to comment on the evidence itself at all because if I did I would actually have to offer quite a bit of details and that’s just too time-consuming (frustrating statement, I know). But let me just say that if you look at the relevant passages in modern grammars of Koine, you’ll notice that Carrier does not do justice to the article, nor does he actually get the nature of the exceptive clause. He doesn’t even seem to have understood the issue in the secondary literature (i.e. what is confusing and what is not). But I don’t want to go into more detail concerning the lack of interaction with secondary literature. What I meant is something like this statement: “Certainly in Gal. 1.19 Paul meant either James the Pillar or another James. And if he meant James the Pillar, then he did not mean he was literally the brother of Jesus—as that James appears to have been the brother of John, not Jesus.” The idea that James the Pillar might be James, the brother of John (i.e. son of Zebedee), is something that isn’t discussed in commentaries for the reason of the wide consensus, that this James had already died at that point in time. (For a good overview over some of the relevant issues concerning James, I’d suggest you consult Bockmuehl’s “James, Israel and Antioch”.) If a student of mine wrote a sentence like the one I’ve just quoted, I’d have the suspicion that he or she has failed to understand the secondary literature on the subject. In other words: you just can’t write on the possibility that one of possibly two James is the brother of John without there also discussing the consensus view that he had already died at that point in time. But that’s just a very minor and random observation on Carrier’s interaction with the secondary literature – even though it’s the kind of problematic interaction that I would not accept from a student. But again: not substantial at all compared to Carrier’s much deeper problem (see below). It might be the main reason why many of my colleagues don’t think his work can be taken seriously – and in consequence that Bayes’s theorem too is of no value for historical studies – and it’s indeed below any standard of a publishable contribution to biblical studies, but it’s not at all the biggest problem I have with Carrier.

“Understood. In one sense, though, and I suspect you would agree, Bayesian reasoning is not really a ‘new method’ at all but rather a symbolic representation of valid reasoning processes at their best.”

Absolutely. Perfectly put. But it might have the appearance to others and so I try to emphasise what you say here while also pre-emptively reacting to such a position. Sometimes (!), Bayes’s theorem can also cover invalid reasoning, however. Our brains have difficulties to follow Bayesian lines of thought in real time and in the past I’ve made mistakes in applying the theorem myself in presentations before people with degrees in Maths and they didn’t even notice. So I think it’s great to have Bayes’s theorem in mind when doing research but I think one should be very careful how one incorporates it in the actual written product, in that it can also be a distraction. I think that’s exactly what happened in Carrier’s case. The way he updates the priors is not legitimate because of the way he estimates the likelihoods. But that can easily be overlooked. The danger with reference to Bayes is that whenever someone says “and here’s my likelihood” and “here’s my prior” we are so occupied with thinking about whether the calculation behind that number might be correct that we can miss the fact that the calculation itself might be coherent but simply not leading to the goal in question (see below).

“Carrier does not argue that the Greek expression Ἰάκωβον τὸν ἀδελφὸν τοῦ κυρίου itself favours mythicism.”

I was quite deliberate in how I put it, but again it’s possible that the way I expressed myself lent itself to be misunderstood: “To say that this wording might actually “favor” the myth-hypothesis …” I used “might” because he says (as you quote, p. 591): “My own conclusion is that there is at best no difference in probability and at worst a difference favouring myth…” Next page: “I actually think this evidence is twice as likely on mythicism…” For the purpose of my comment I didn’t care about how in the end he factors it in, what I found shocking was even the fact that someone with a basic knowledge of Greek and Bayes might think the favouring might go in this direction. It clearly doesn’t and it not even does so on the basis of Carrier’s own interpretation of the other passages (on this below).
“Again, having had another quick look at Carrier’s argument here I don’t believe he claims the actual ‘word choice’ or phrase itself is the grounds for his argument about its significance. (Carrier does address other Greek phrases found in Paul to support his argument, by the way.)”

“I may be mistaken but it appears to me that your view of Carrier’s argument is limited to his final paragraph which he begins: So the question at hand is how likely it is that Paul would use the phrase ‘brothers of the Lord’ on the two occasions he does . . . . (p. 591) But the context of the preceding argument (587ff) is clearly about far more than the simple wording or translation of the phrase “brother of the Lord”.”

Well, the whole section clearly is about the fact “that twice Paul mentions ‘brothers of the Lord’” (p. 582). So we are dealing here with a piece of evidence – an expression, a genitive construction with two specific nouns – and the question of which of two hypothesis it favours. This classical Bayesian confirmation. I’ve done that for a single word θριαμβεύειν (and it took me a whole monograph). Nothing easier than that. However, all Carrier does is to argue that it was indeed possible to use this expression to refer to fellow Christians without a physical link to a person called Jesus. Now, let me be even clear: even if of the two passages he discusses in detail, the expression would clearly refer to this latter category (which I think is not the case), this doesn’t tell us anything about P (“brother[s] of the lord” I reference to believers in general). That’s also wrong in the blogpost you linked to, by the way (“So how expected is the evidence if our hypothesis were not true? I have to say it is “extremely probable”: that is, 0.99.”). The ratio of attestations of different meanings of one lexeme/phrase doesn’t tell us anything about the likelihood. Or, rather, at best, a single attestation proves that the likelihood can’t be zero. So it tells us something (but it’s never zero anyway because of how language works anyway, so that’s not such a great insight). But it certainly doesn’t tell us how the likelihood-ratio looks like. It’s entirely possible that most occurrences of the phrase “brother of X” means ‘fellow Christian’ and not ‘son of the same mother’ but the likelihood still being clearly in favour of the latter! For example: the verb θριαμβεύειν almost always or always occurs with a specific meaning X. But that doesn’t mean that P (θριαμβεύειν I X) = 1! Rather, I have to look at the concept of X and how it is usually realised lexically. On that basis I might find that the likelihood is more like 0.2 or so because there are other lexical solutions that are more “expected” (such as πομπεύειν εν τω θ. κτλ.). The expression still favours interpretation X because the likelihood of θριαμβεύειν given interpretations Y and Z is even lower because there are many much better ways of expressing that thought (αγειν θ. κατά τινα κτλ.). The fact is: I can’t know unless I look at the frequency of these alternative word choices. If I don’t want to do that, I just can’t comment on the likelihood. It’s outright impossible.

So in order to determine the two likelihoods Carrier is interested in he would have to ask the two questions:
(1) How can one refer to fellow believers (“those in Christ,” “those of the faith, etc.”)? And how frequent are these solutions?
(2) How can one refer to physical relatives (“the other son of mother Mary, etc.”)? And how frequent are these solutions?

Unless one has done this very hard work, it’s simply impossible to estimate likelihoods. Everything else is fundamentally mistaken. So I actually don’t care whether in the end – after examinations of the different ways to say ‘believer’ and ‘physical relative’ – the ratio of likelihood turns out to be 2:1, 1:2, 10:1 (as you suggest) or 1000:1 (which seems much more realistic to me on the basis of my reading of ancient Greek texts). The basic problem is that it’s logically impossible to come up with an estimation of the likelihood on the basis of the evidence that Carrier discusses. That’s the kind of problem that I mentioned above that might come from relying too much on Bayes, i.e. trusting that it will guide us to the right result: if we make a mistake in applying the different factors, the whole things just blows up. That happens, to be sure. In fact, Carrier’s mistake is one my students like to make when they are asked to apply Bayes’s theorem for the first time. It’s not a big deal. But it shouldn’t be published and it can’t be taken seriously as a scholarly argument in my opinion.

Ok, I hope this helps a bit. Please understand that I can’t comment on this any further due to lack of time. It’s not a cheap excuse. I do feel bad about it because I like the way you approach these issues. If my comment is still unclear to you, perhaps you might profit from taking a look at “Paul’s Triumph.” There you can see a demonstration of how a likelihood can be estimated (though I refrain from using numbers even in this well-studied example) and what Carrier would have had to do in order to come up with his estimations in my opinion.

Still more on the page that follows…


Browse Our Archives