I'll ask you what I asked him, but open it up a bit to make it easier to answer Do you have any evidence, argument or demonstrative proof that the Bayesian approach to uncertain inference performs less well than any method of your choice when the available data are of low "quality," meagre quaniity or of nearly equivocal bearing?
That wasn't really my point there, though.
My point was that the data often available in ancient history is too limited and unique for probability functions such as BT, especially BT, if we are discussing Historicity.
The way BT works is by having multiple data sets to input, however, Historicity in Ancient History is often incapable of delivering sufficient volumes of data to feed into for comparison of a phenomenon against the normative.
There is often no normative to compare against.
Using BT on such results in what Carrier did, running the phenomenon against the logic derived from the phenomenon, or, running the data against data not of its kind, but in portion s isolated from the whole.
To use BT on Jesus, for example, we need another figure like in form of textual culture and Ancient record, yet one known to be existent, and would better be served by two or more such examples.
The only real other example is Buddha, but Buddha is in the exact position as Jesus as to the quality of Historicity, so whichever case we assign Buddha, so then would BT likely agree, but that is even poor, for that is only one comparison, where truly we would be more accurate with 3 or even 4 such examples to compare Jesus against, so that our question is accurately "does this figure, of whom several sects and texts arose around, with little evidence otherwise, stand likely to have existed?", as that is the actual question since all inferences from the data about the figure and culture, feeding into BT for comparison against "what people and life was like in ancient X location", is inherently going to cause error as we just assumed that the text is accurate about who and where within the text when no such guarantee is granted, snd while such may be small issue to the Historical method, it is rather s huge issue in BT probability, as it greatly alters what we are actually asking.
As such we are no longer asking if the phenomenon indicates a real person, but are asking if the text matches real expectations of the era we assume it to be addressing.
No, the texts are clearly off on several factors regarding the era we assume them to be discussing; hell, we can't even be certain we have the era right and have to accept a very wide birth of era possibility since the texts discuss things ranging from 1 c BCE to mid to late 1 c CE.
And this is the problem with BT in general for historicity.
Now, though it is slim, if we assume the standard view that Joseph Smith's accounts are not real outside of his mind or account, then we could use BT to compare Joseph Smith to Paul's vision and see the probability that Paul's vision did not occur in reality, but this is even slim for comparison as it is 1 to 1 comparison, where BT is best as applied by Turing and we have hundreds of data bits to enter into the several passes for computation.
One pass and one piece of data back against itself, or against only one other data is going to produce a very limited probability quantification.
It does not suffice to say we'll just add more later, for that is not only unlikely, but accepting multitudes of "it didn't happen" from a position which openly admits it lacks far more information than it will ever be capable of finding.