Perhaps the best
thing that’s ever happened to slippery grifter and unpleasant manipulator
Richard Carrier on the Internet is my set of articles in which I try to help
him reason better, be more honest, and become a kinder person (see here, here, here and here). But, alas, he got totally
triggered and decided to write a horrendous dismissal of my critique of his
terrible article on using Aristotle’s existence to undermine Jesus’ existence.
Now, I normally
wouldn’t waste this sort of time on discredited people like Carrier – he’s the
young earth creationist of the world of mythicist history. But as regulars will
know, I do use the exposing of fraudulence and bad reasoning to showcase deeper
psychological and philosophical tools at play in how masses of people can be
manipulated, in the hope that it will help others see through the cheat. So,
blog posts like these can serve as useful tools in aiding readers in critical
evaluation and spotting bad arguments, even when they are - and perhaps especially
when they are - presented with the outward appearance of sophistication.
Carrier’s modus
operandi is a classic case in point – and while he’s not taken seriously by
virtually all credible professionals in any of the subjects about which he
writes extensively, he does seem to have assembled a small micro-cult following
from a few thousand people who he’s duped into thinking he’s someone worth
following because he’s onto things that even most experts have missed. Spoiler
alert: he’s not; he’s a crank, and, from my personal experience, a deeply
unpleasant one at that – which I’m going to expose, in a moment by showing how
he manipulates the subject matter to sound like he’s offering sound
refutations, but first, on the psychology of how his readers could possibly
fall for such a blatantly untrustworthy front man.
I don’t care if
Carrier reads this – he is almost certainly never going to change. But I do
want his readers to read this, because they deserve better, and should want so
much better for themselves, just as I want so much better for them. I’ll use
Carrier’s own words from his article criticising me to show you every element
of his deception, in the hope that it will help his readers see through him,
and in the hope that it will give my readers a lesson more generally in how to
see through people like Carrier.
Because the first
thing that makes people like Carrier interesting for all the wrong reasons is
how it can be the case that his readers can’t see through him, when it
should be so much more obvious to them that he’s not to be trusted.
As a case in point,
let me state right here that I’m going to submit this on the comments section
of his blog post, but I’ll bet you don’t see it appear – because he won’t allow
anything in the comments that shows him up at this level.
Before I get to
Carrier’s errors of reasoning and mistaken arguments, I’ll first expose the
primary psychological cheats he employs.
1) The dismissive
insult
Carrier’s tactic of
insulting dismissals, the one he employs most often (“C.S. Lewis may have been
the worst philosopher of the twentieth century”, “AI Is garbage“, “Jordan
Peterson is a crank”, any expert who disagrees with him is a ‘buffoon’,
‘doofus’, ‘bozo’ etc ), is a deliberate strategy employed by many manipulators
of his nature. Because Carrier’s worst fear is being exposed, his ‘go to’
strategy is to try to intimidate readers and critics into thinking disagreement
= stupidity. It’s an attempt at psychological dominance and frame control,
where his own followers are made to feel judged for even contemplating someone
is onto him, and encouraged to dismiss not just all the experts who have
exposed his work as wrong and his arguments as sub-standard, but be dissuaded
from checking it out for themselves. That is what cult leaders do – it’s as old
as the hills.
Of course, it can
only work on people either too uniformed or too malleable to spot the cheat.
I’ll give you an example which appears in literally the first thing he says in
introducing his critique of me; he says “Today I make an example of an internet
buffoon”. Now wait. Just stop and ask yourself several things about that. Any
of his readers could look at my blog and find exactly who I am. I’m someone
with over 1000 blog posts on the Philosophical Muser site to which he linked,
and hundreds of articles of mine for other institutions. In that body of work
you would find some of the most comprehensive and consistent in depth analyses
of most subjects out there – there is no way anyone with half a brain could
look at my material and write me off an as ‘Internet
buffoon’. Moreover, Carrier knows I’m not an ‘Internet buffoon’, so why did
he lie and say I am? In other words, why is the very first thing he said a lie?
It’s easy, because unless he lies, he can’t get his readers on side, and by starting
with a blatant lie he can lay the foundation down to have them fall for his
nonsense reply to my article.
Think about what
that means if you’re one of his readers – he just lied to you, knowing that if
you searched through my work you’d know he’s lying because you’d know I’m not
an ‘Internet buffoon’ – I’m actually much more on top of these subjects than
Carrier (which is why he refused to debate me live on video – see below). But
it also shows something else telling; it shows that Carrier has contempt for
his readers – he doesn’t care that you could find out the truth really easily –
he’s just expecting you to swallow it without care because he has so little
regard for you. And if you think about it, that’s exactly what you’d expect. He
can’t really value his readers beyond your merely as serving as props for his
arch manipulations, because he thinks most of you are so gullible you’ll just
swallow it up hook, line and sinker. How can he respect people who he knows
fall for his cunning deceptions, and thank him for it afterwards?
It reminds me a bit
of the Nigerian prince scam, where virtually nobody falls for it, but the
scammers still use the Nigerian prince method on the basis that if you're one
of the few gullible people left in the world who hasn't been apprised of the
Nigeria scam and are likely to fall for it, you will be of those for whom the
uniformity of the 'Nigeria' email won't be alarming and prohibitive. With the
simplicity of using Nigeria each time, the scammers save having to waste time
with all the false positives, and they will continue to catch in their net the
few gullible fish still in the sea. Carrier has contempt for his followers
because the fact that you’ve become one of his followers has already showed him
how gullible you are. As said, I want so much better for you.
By the time he’s
moved on to his next line in the article, he’s already told a second big lie:
saying my private messages to him reflect “someone painfully immature and
emotionally irrational, with no interest in learning or understanding or
engaging in any kind of intellectual discussion”. Well, yes, they do – but they
reflect those things about Carrier, ironically – and I have screen shots of
them to back that up if ever needed. I won’t labour on that point, but suffice
to say, Carrier wasn’t willing to take up my offer of a live debate – where he
knows there’s nowhere to hide and where the fraudster is forced to think on
their feet, and can’t hide behind carefully manipulated, long-winded written
texts – he showed a firmer desire to have a written debate, but when I left a
voice message to explore this, he said, ‘I don’t listen to voice messages’
(which is ridiculous in this context, but expected from a narcissist like him)
– and then when I sent him a written message giving some suggestions about what
we could debate, he simply brushed it off with ‘tl;dr’. He’s an absolute joke
of a man.
I mentioned that he’s
a narcissist – which you can probably sense by his own words on his Facebook
profile:
Yes, I'm the real
Richard Carrier. Sadly, I'm much too busy to hang out here much. But I have to
maintain a FaceBook page to prevent malicious hacks from pretending to be me.
What a scream that
is – compounded by the fact that most narcissists have no idea how deep their
own narcissism runs. Virtually no one outside his cult knows who Richard
Carrier is, yet he has the audacity to assert “Yes, I'm the real Richard
Carrier”, who’s “much too busy” to hang out with the likes of you lot, but I’d better
have a page to “prevent people from pretending to be me”. Haha, I honestly can’t
think of a single scenario in the world where someone would have the even a
flicker of inclination to pretend to be Richard Carrier. Hilarious.
2) Deflection
through projection
Now, let me
show you another tactic employed my manipulators like Carrier – he chooses
terms of insult, like ‘childish’, to shift attention away from his deceptive
methods and onto a manufactured image of his critic as the villain. It’s one of
the oldest psychological tricks that insecure thinkers like him play. He wants
you to think I’m childish so you don’t take my criticisms seriously - in fact,
he uses this accusation that I’m childish EIGHTEEN times in the article, no
less. Again, it’s another fat lie that shows utter contempt for you, his
readers, because you can quite easily find out for yourself that I’m not a
childish person. What I do have, as evidenced from highlights of what I sent
Carrier in messenger over the past few months, is the measure of him, and a
desire that he no longer gets away with his manipulations online. That is the
opposite of childishness. Here are some examples of what I sent to him:
I've looked
at some of your writings today, and two of the primary things that stand out
are this:
1) You
pretty much always conjure up in your head what you want the conclusion to be,
and then use any squalid method possible to argue for that.
2) Your body
of work is similar to that of a young earth creationist, except with the signs
reversed. You leave out so much of the vital material, you distort the
arguments of your opponents, you argue dishonestly, and your operate on
counterfactual peripheries in the hope of forging a niche reputation, where any
expert in the field would easily show your ideas to be wrong, and you produce material that relies on the ignorance
and pliability of your audience, where your only recognition comes from people
too intellectually lightweight to see through you.
Your
writings don’t come across that way – the most important thing to you is how
others perceive you, and your asset or non-assent to truth is driven by how you
think you can get more popular and regarded, which means claiming to hold
positions that are niche enough to give you a certain type of following. You
also claim to care what people who disagree with you think, but that is not
backed up by your policy of limiting comments on your page/articles – which is
exactly what one would expect of someone about whom the above agenda applies.
Have you not
noticed that you always write off people as cranks, when you're the crank? Have
you also noticed that cranks, like YECs, almost never do live debates either,
because they don't want to be exposed?
Even though
I was being friendly, the reason you reacted unpleasantly is because all
narcissists do this when they are deeply insecure and have been found out by
someone who has the measure of them. Biting back aggressively is a classic sign
of being found out - it’s what Waugh referred to in one of his letters as “the
concealed malice of the underdog”. All that you said in your hostile response
is undermined by the fact that you don’t have the courage or intellect to
debate me live, where you would have to display your mind, think quickly on
your feet, and not hide behind walls of manipulative text within the cosy
self-incarceration of your echo chamber.
What you do, sir, is not truthseeking, and nor does it show good
character – it’s a cliched Promethean cheat that will only bring you shallow
and transitory pleasures – not ultimate joy, fulfilment, contentment or wisdom.
When you reach the age of realisation of this, I’ll be happy to have a
civilised conversation with you.
The fact
that you dismissed all that in a few minutes shows me you can't be taken
seriously, and just want to believe what's convenient for you.
And the same
with all the attention-seeking about being godless, polyamorous, etc. The big
contradiction at the heart of polyamory, I think, is that a person who is not
mature and cultivated enough to comprehend how their needs can be fulfilled in
the right monogamous relationship with one beloved, will also not be mature and
cultivated enough to realise that their needs cannot be fulfilled through
multiple relationships or sexual unions, however plentiful the number and
varied the experiences. Things can still be much better for you, but not until
you change and become wiser. And if those comments induce rage in you, that's
because you know how much force it takes to defend a fragile house of cards.
See what I
mean? Not childish at all, is it? It’s hard-hitting, sure, but it’s
hard-hitting because people like Carrier need to be told that they won’t always
get away with things. And there’s nothing in any of my public works that is
childish or immature. It’s only what Carrier needs you to think to put the
guard up against his being exposed. Carrier, on the other hand, does appear
childish in virtually all his conduct – private or public. He is the one who
writes off people with crass dismissiveness; peppers his blog posts with
puerile Americanisms like ‘doofus’ and bozo’; isn’t even mature enough to
sustain a faithful, monogamous relationship, and insults anyone who doesn’t let
him get his own way. That is the epitome of childishness.
The arguments
Let’s look further
now at his attempted rebuttals, because they serve as a good lesson in how to
‘sound’ clever and competent when really it’s a litany of further errors of
reasoning, bad arguments, and YEC-like distortions of data and fact. His
attempted rebuttal fails at every critical level the first article does, but to
an even greater extent, because this time he is trying harder to dig himself
out of a bigger hole.
He still rigs his
Bayesian model from the outset by inventing reference classes that already
assume Jesus is fictional, still assigns precise-looking likelihood ratios that
are wholly uncalibrated, empirically ungrounded, and indistinguishable from
personal intuition dressed up as mathematics; he still compounds this by
multiplying dependent, culturally and textually interconnected evidence streams
as if they were probabilistically independent, grotesquely inflating his
results in violation of basic Bayesian requirements; he still, in classic
Carrier style, responds to criticism by redefining dissent as ignorance of
Bayes; and yet again, his ultimate proof of failure is that his method reliably
produces fringe conclusions rejected by the historical discipline at large, not
because historians fear probability theory, but because his model fundamentally
misunderstands how historical evidence, sources, and uncertainty actually work.
To save wasted time
going through dozens of his errors, I will just pick the ones I think come
across as his best attempts at rebuttals – which are still sub-standard – but
they are the ones that I think will most fool his readers, because on the
surface they look like the most sophisticated. Again, I’m doing this to expose
the subtle art of squalid manipulation, and show how it can easily come across
as highly sophisticated, intellectual criticism. We’ll start with his comments
first, then my comments underneath.
Richard Carrier 1#: Muser confuses the fact that we have evidence specific for Alexander
the Great that keeps his posterior probability high with meaning his prior
probability should also be high. It’s the other way around: if all we had were
the wild myths of Alexander, his prior and posterior probabilities would be
low, and all historians would agree—as they do for Hercules and Osiris, for
example. The only reason we conclude Alexander existed is because of the
evidence specific to him. Therefore, mathematically, his prior probability,
which is the probability prior to considering all that evidence, cannot be
affected by that fact. It therefore would not go up. It would stay low. It
would remain in fact exactly what it would be if that evidence didn’t exist. That’s literally the
point of the prior probability.
My comment: Oh wow, boy is this man slippery
in his confusion. Most extreme people online either speak cleverly and
competently or they speak foolishly and incompetently – whereas Carrier is one
of those rare beasts who tries to perfect the art of sounding clever and
competent while misleading you with hollow claims. I believe that’s why he
draws in a certain unsuspecting following. Carrier’s claim sounds confident,
but it is conceptually mistaken, and the mistake reveals exactly where his
Bayesian reasoning departs from how priors actually function in historical
inference. It’s basic stuff; a prior is not “what you would believe if you
erased all relevant background knowledge and pretended only myths existed.”
That is not what a prior is in Bayesian reasoning, and it is certainly not how
priors are set in any applied field, including history. A prior is the
probability of a hypothesis before considering the specific evidence under
adjudication, given all relevant background knowledge. Background knowledge is
not “the evidence being tested,” but the wider contextual facts that frame what
kinds of hypotheses are plausible in the first place.
Carrier’s
hypothetical - “if all we had were the wild myths of Alexander” – is a high
school category error. If all we had were myths, then those myths would
themselves become part of the evidence under evaluation, and the background
knowledge would change radically. In that imaginary world, historians would
also lack knowledge of Macedonian kingship, successor states, administrative
continuity, geopolitical causation, and the normal sociological patterns of
royal propaganda. That is not a legitimate way to define a prior; it is a
counterfactual that strips away the very background conditions that inform
rational priors.
For
Alexander, the prior is higher than for Jesus not because of later literary
attestation, but because the background facts already make his non-existence
implausible (that is, empires do not appear without founders, successor states
do not arise from fictional conquerors, and administrative and political
continuity does not emerge from pure myth, that sort of thing). Those facts
belong in the prior because they are not the disputed evidence; they are the
historical context that constrains what explanations are reasonable. Ignoring
them is methodological carelessness.
Moreover,
Carrier’s statement that “that’s literally the point of the prior” reverses the
concept. The point of a prior is not to simulate epistemic ignorance; it is to
encode justified background expectations before updating on contested evidence.
The problem is not that Carrier uses priors - obviously; it is that he defines
them in a way that excludes precisely the kind of background knowledge
historians are obligated to use. If you Google research Carrier’s historical
disputes with others, you’ll see this plays out repeatedly in his tactics. And
please do the research and see for yourself.
Carrier’s
other repeated mistake here is the category error of treating priors as if they
must be set in a historical vacuum stripped of all contextual background
knowledge except “myth vs non-myth.” That is not how priors work in applied
Bayesian reasoning, and it is certainly not how historians reason. A prior is
not “what you would think if you pretended no contextual evidence existed at
all”; it is what you assign before evaluating the specific evidence under
dispute, given all relevant background knowledge. Carrier’s attempt to freeze
the prior at a myth-only baseline misunderstands that priors legitimately
incorporate historically grounded context, not just genre labels. In short,
Alexander’s prior is higher than Jesus’s not because of the later literary
evidence, but because the background conditions surrounding Macedonian kingship
already make non-existence unlikely.
Richard Carrier 2#: Only when we add that evidence does it go up, but that’s then the
posterior (not the prior) probability; and all of the change from the one to
the other is in the likelihood ratio, not the prior. Alexander’s prior stays
the same as for Jesus, which I set at 1 in 3 in my article on Aristotle, same
as my original study (it has since been reduced to 1 in 4, as I there explain,
but I set that aside to argue a fortiori). That Muser doesn’t know this shows
he has no actual understanding of Bayesian reasoning or mathematics, and thus
is not at all qualified to critique any application of it.
My comment: Look how he uses forceful
rhetoric to make a statement that laypeople probably wouldn’t spot is just flat
out conceptually wrong. Carrier is correct about one narrow point of formalism
– that likelihood ratios are what update priors. But he is flatly wrong to
insist that Alexander’s prior “stays the same as for Jesus.” That claim only
follows if one defines priors in an artificially stripped-down way that
excludes historically relevant background knowledge. Bayesian reasoning does
not require identical priors across hypotheses; it requires justified priors
given what is already known before examining the specific evidence at issue.
The crucial
mistake is that Carrier treats the prior as though it must be set before any
historically meaningful differentiation is allowed. But priors are not
epistemic blank slates. In historical inference, they legitimately encode
background asymmetries such as political context, sociological plausibility,
institutional continuity, causal footprint, and what have you. Alexander is
embedded in a well-understood class of phenomena - Hellenistic kings, imperial
conquest, succession crises, and administrative aftermath - that rarely, if
ever, arise from fictional individuals. Jesus, by contrast, belongs to a class
of itinerant religious preachers in a highly myth-making environment where
symbolic invention and theological personification are known mechanisms. These
asymmetries belong in the prior, not forcibly suppressed until the likelihood
stage. The fact that Jesus is so well attested is what shifts the posterior,
but it does not retroactively inflate his prior; the background context keeps
his prior lower than Alexander’s, even before examining the evidence.
By insisting
that all differentiation must occur through likelihood ratios, Carrier is not
following Bayes neutrally; he is imposing a modelling constraint that
artificially equalises starting positions in order to push all discrimination
into later stages where he can manipulate likelihood assignments. That is not honest
work.
Finally,
Carrier’s accusation that disagreement here shows “no understanding of
mathematics” is equally dishonest. The objection is how he creates the
parameter. One can fully understand Bayesian updating and still reject a model
that defines priors in a way that deliberately excludes relevant background
knowledge. That is methodological disagreement – and the fact that he thinks
it’s a mathematical disagreement shows he doesn’t really understand what is
being debated here, never mind where his errors lie.
In fact, I’m
probably being too charitable still – because this is where Carrier’s Bayesian
formalism detaches completely from historical reality. He is correct
mathematically that updating happens via likelihood ratios - but he is wrong to
insist that priors must therefore be equalised across radically different
historical contexts. Bayesian reasoning requires justified priors, not
identical priors. Treating Jesus and Alexander as starting with the same 1-in-3
prior simply because both later acquire mythic narratives ignores massive
asymmetries in background knowledge: political structures, documentary
cultures, social incentives for invention, administrative continuity, and the
presence or absence of motive and mechanism for fabrication. Bayesian models that
flatten these distinctions are not neutral; they are distorted. Carrier’s
accusation that disagreement here shows “no understanding of Bayes” confuses what
I would call mastery of a formula with competence in model construction. He
can’t see this because he’s not on top of this subject. The sleight of hand
trick is in deliberately stripping priors of historically relevant information
in order to force all differentiation into the likelihood stage – well, I say ‘sleight of hand trick’, but with Carrier it’s not
always evident whether he’s being deceptive or just confused, because he is
both so often.
That is not
principled Bayesianism; it is model engineering - and I see the same thing with
so many other guru figures that the question of honest ignorance or deliberate
deception continually looms large. Carrier is more like the equivalent of YEC’s
Grady McMurtry than Ken Ham. I’m convinced that Ken Ham is a simpleton as well
as dishonest. I’m not sure Grady McMurtry is that simple – but he’s cunningly
deceptive, and Carrier strikes me as more like the latter.
Richard Carrier 3#: This is also demonstrated by the fact that he didn’t know my
likelihoods for Aristotle not existing already take into account that
evidence’s dependency on “a Peripatetic tradition.” He didn’t even know that
this is always the case, that it is literally what you are doing when you
estimate likelihoods in Bayesian reasoning: you assume the one condition is
true (“Aristotle existed”) and then assess how likely the evidence would be
dependent on that being the case; then you assume the other condition is true
(“it was just a Peripatetic tradition”) and assess how likely the evidence
would be dependent on that being the case; and it is the ratio between those
two likelihoods that tells you whether that evidence argues for or against
either condition (or neither), and the size of that ratio (the difference
between those two likelihoods) tells you how weak or strong that evidence is
for that conclusion.
My comment: Here Carrier
confuses naming dependence with modelling it. While he correctly describes the
formal definition of a likelihood ratio, merely imagining an alternative
hypothesis (“it was just a Peripatetic tradition”) does not automatically
control for evidential dependence. When multiple pieces of evidence arise from
the same causal pipeline, they are correlated, and multiplying likelihood
ratios still double-counts information unless that correlation is explicitly
modelled. Carrier does not do this. He assigns separate likelihood ratios to
Aristotle’s writings, testimonies, inscriptions, and later historians, and
multiplies them anyway; reducing the numbers does not fix the inflation. Worse,
he applies this leniency asymmetrically: dependence is “handled” for Aristotle
but used to collapse multiple Jesus traditions into near-total dependence.
Again, is it confusion or dishonesty? Either way, it is not principled Bayesian
reasoning; it is informal judgment expressed with the appearance of rigour but
none of the control. The issue is not whether he conceptually acknowledges dependency, but
whether his numerical assignments properly discount correlated evidence
streams. In the Aristotle case, nearly all the evidence flows through the same
Peripatetic–Hellenistic transmission networks. Treating each as yielding an
independent likelihood ratio (even a reduced one) still double-counts shared
provenance unless explicitly modelled as conditional on the same causal chain.
Carrier never provides such a model; he simply assigns smaller numbers and
multiplies anyway. That is not Bayesian rigour, and his readers don’t spot
this. Worse, he applies the exact opposite standard to Jesus, collapsing
multiple traditions into near-total dependence even when they demonstrably
diverge in theology, chronology, and Biblical setting. This asymmetry shows him
up for what he is - that he has an agenda, and that his “dependency adjustment”
is not principled, it is outcome-driven (which is exactly what my first article
expressed too).
Richard Carrier 4#: Every probability in every application of Bayes’ Theorem is a dependent
probability. And item by item you can run strict dependency in likelihoods by
iterating for each item of evidence in chronological order. But when this has
no effect on the outcome, there is no reason to do it, because it requires
needlessly complex arithmetic, to no improved result. There is only a
dependency on each hypothesis, which is already baked into Bayes’ Theorem. That
is precisely the dependency relation we are calculating when we derive the
likelihood for each hypothesis: the probability of that evidence dependent on
each hypothesis being true—and dependent on all human background knowledge,
because each likelihood is P(e|h.b), where the probability of e is dependent on
the truth of h and the entire contents of b (such as all inscriptions from
Greco-Roman antiquity and their empirically observed frequency of forgery).
This is Bayes’ Theorem.
This no dependency on “a Peripatetic tradition” has
any effect on this estimate, because the question is whether Aristoxenus is
fake or a liar, not where he got his notions from. On Bayesian reasoning we are
assuming Aristoxenus got it from just ‘a Paripatetic tradition’ on the contrary
hypothesis, and so we are already accounting for its dependency on that
tradition. Our likelihood is a dependent probability in that sense, because it
has to be: that’s what Bayes’ Theorem is doing. We then put that in ratio to
the other hypothesis, where we assume Aristoxenus got it from being an
eyewitness partner of Aristotle, and asking how likely the text we have is
then. Those two probabilities then sit in a ratio, to give us the weight this
evidence has in attesting that Aristotle existed or not.
My comment: Again, there are real,
substantive mistakes here, not in the algebra of Bayes’ theorem (though I don’t
know if that’s a mere fluke on Carrier’s part), but in how dependence,
background knowledge, and evidence aggregation are being handled. At first
glance, Carrier doesn’t appear to make any mistakes about Bayes’ theorem in the
abstract in this section, but he misapplies it in ways that matter.
Here’s where
it goes wrong. He equivocates between “conditional on a hypothesis” and
“independent pieces of evidence”. It is true that every likelihood is
conditional as per the correct formula, but that does not mean that multiple
pieces of evidence are automatically independent given the hypothesis.
Conditional dependence between evidence items is a separate issue. Bayes does
not magically “bake in” evidential correlations. And when Carrier says “You can ignore
dependency if it doesn’t change the outcome” - that is methodologically
backwards. Whether dependency changes the outcome is exactly what must be
shown, not asserted. You can’t justify skipping dependency modelling by
claiming in advance that it won’t matter - especially when the central question
is whether the evidence represents multiple independent attestations or
repeated recycling of the same source. In Bayesian terms, this is equivalent to
assuming conditional independence because it is convenient, not because it has
been demonstrated.
Carrier is
right that likelihoods are conditioned on background knowledge. But bundling
things like inscription frequency, literary survival, institutional
transmission, and genre practice into a sloppy group does not solve dependence
- it obscures it. Background knowledge constrains plausibility; it does not
license treating correlated signals as separate evidential hits. Carrier
actually grossly misstates what the dependency objection actually is. The issue
is not merely where Aristoxenus got his information, but whether multiple
attestations reduce to a small number of causal sources. If they do, then each
new item adds far less information than Carrier’s multiplication assumes.
Saying “we already assume tradition on the null hypothesis” does not address
whether the same tradition is being counted multiple times as if it were
independent confirmation. And as I said above, he also applies dependency
asymmetrically, especially when Jesus comes into the equation.
Conditioning
on a hypothesis is not the same thing as modelling correlation between evidence
items, and Bayes’ theorem does not excuse you from that work. Bayes’ theorem
does not license multiplying likelihood ratios just because each likelihood is
conditioned on the hypothesis. If you could see this, you’d see that Carrier’s
“smaller numbers” defense is mathematically irrelevant. Reducing likelihood
ratios does not solve dependence. Whether the multiplier is 100:1 or 2:1,
multiplying correlated evidence still inflates confidence. Or to put it another
way, dependency is structural, not scalar. You fix it by modelling joint
likelihoods or collapsing evidence streams - not by eyeballing smaller ratios. If
you are unsure who to believe, simply show this blog post to your smartest friend,
those who are most apprised of these subjects, and I’ll wager you’ll hear them
agree with me. Or do the research yourself – if you do it honestly and pay
close attention, you will come to the same conclusion.
You’ll also
see that Carrier conflates “acknowledging dependence” with “accounting for it.”
Carrier repeatedly says he “assumes tradition” on the null hypothesis and
therefore has already handled dependence. But that only specifies a story about
causation; it does not mathematically prevent multiple evidence items generated
by that same story from being counted repeatedly. Bayesian rigour requires
showing how much new information each item contributes given the others.
Carrier does not do that. Again, I don’t know if it’s dishonesty or deception,
but Carrier aggressively collapses Jesus traditions into near-total dependence
while allowing Aristotle’s evidence - much of it flowing through the same
educational, textual, and institutional channels I mentioned in my first Blog
post - to count as multiple semi-independent hits. That is a discretionary
modelling choice that should fool no one.
Carrier seems
to understand that all probabilities are conditional - but he is damn wrong to
imply that this dissolves evidential correlation. Bayes’ theorem requires you
to deal with dependence; it does not excuse you from it. Treating “background
knowledge” as a black box does not magically absorb correlated causal
structure. Carrier seems to comprehend a bit of Bayes’ formula, but he misuses
it by treating correlated historical evidence as quasi-independent without a
formal dependency model, then defending the move rhetorically rather than
mathematically.
Richard Carrier 5#: This is an equivocation fallacy, where Muser has switched illicitly
between two different meanings of “mythologized,” and ignored the actual
data-driven process he is incorrectly describing. First “therefore Jesus is
mythologized” isn’t a statement that he didn’t exist (because historical people
are also mythologized). So “Jesus is mythologised therefore Jesus is
mythologized” is not a circular argument but a tautology. It’s as true as
“Jesus is a man, therefore Jesus is a man.” That Jesus is heavily (not merely)
mythologized (unlike Aristotle) is a fact (not an argument capable of being
circular). It is a thoroughly, professionally documented fact. The consequences
of that fact are that Jesus starts out as likely to exist as anyone else
heavily (not merely) mythologized (unlike Aristotle), which is empirically
observed to be no more than 1 in 3 times. Muser is thus just not paying
attention. He doesn’t care about what I am actually saying or referencing in my
article. He wants to emotionally rewrite reality so that 1 in 3 is “very low”
when in fact it’s generously high, that this predetermines the outcome when it
routinely doesn’t (and indeed couldn’t), that heavily mythologized heroes exist
just as often as mundane people when obviously they don’t, and that an
empirically documented frequency is a “circular argument” and “not grounded in
history,” all while completely ignoring (and dishonestly failing to inform his
readers of) what I actually said, argued, and demonstrated professionally with
evidence.
The likelihood of a certain cause (e.g. a mere
legendary person, or a historical person, the two hypotheses being predictively
compared against the evidence here) derives from the background facts of the
frequency of such causes producing such effects. Real revered scientists always
produce more or less this kind of evidence (even if medieval selection
destroyed most of it). Merely legendary scientists almost never do (the
frequency of that in the available massive datasets is not always zero, but is
definitely low). Muser does not appear to understand what likelihoods measure,
or how they derive from extensive field-generated background evidence of what
has typically happened in the relevant period.
My comment: No, no, no. Carrier’s reply
dodges the real objection by sloppily redefining it. The charge is not that
“mythologised” means non-existent, but that degree of mythologisation is being
used as a proxy for existence in a way that quietly bakes the conclusion into
the setup. Calling Jesus “heavily mythologised” is not a neutral fact like
“Jesus is a man”; it is a contested classification whose boundaries depend on
the very historical judgments at issue. Declaring that class to have a fixed
empirical base rate (1 in 3) simply shifts the argument along rather than
dissolving the circularity. Moreover, Carrier conflates frequency with
explanation again: observing that legendary figures “often don’t exist” does
not license treating mythic density as an independent causal variable,
especially when mythologisation itself can be an effect of later reception
rather than evidence about origins. His appeal to “massive datasets” glosses
over genre, transmission, selection effects, survivorship bias, and what have
you - all of which historians normally weigh qualitatively rather than collapse
into a single scalar. Clearly, Carrier has some experience with historical
analyses, so quite why he treats a debatable typology as settled data, then
treats disagreement with that move as confusion about Bayes rather than a
substantive historical critique, is beyond me. Once again, is it confusion or
deception? I suppose even being a seasoned studier of history doesn’t
automatically mean competence in Bayesian analysis.
Richard Carrier 6#: “My reference classes don’t
predetermine the outcome. In that article I am explicit that I use a fortiori
priors that are actually very generous to the alternative, estimating the prior
odds on Aristotle of being historical way lower than is realistic and for Jesus
being way higher than is realistic, to prevent predetermining the outcome. My
exact words (which Muser evidently ignored): “In reality the prior for
Aristotle would be higher (here I am using the most unfavorable lower bound of
the error margin) while for Jesus it would be lower (I am using the most
favorable upper bound … ).” Muser never defends any alternative priors nor
explains how my selection of priors “predetermines” the outcome either way. I
get the impression he doesn’t care if that is even true. He just wants to say
this false thing because he is all emotional about it.”
My comment: No, sorry, not good enough again.
Carrier’s reply completely misses the point. The problem is not how generous he
claims his numerical priors are, but how he defines the reference classes that
generate those priors in the first place. Even a “generous” number is still
methodologically corrupt if it is derived from a category that already assumes
the conclusion. And lowering or raising the numerical values afterward does not
fix this circularity; it merely softens a rigged scale. A prior derived from a
theory-laden classification is still theory-laden, even if its numerical value
is deliberately moderated. Moreover, demanding that critics supply alternative
priors is a red herring; if you really want to postulate such an outlandish
theory that Jesus didn’t exist, the burden is on you, Carrier, to justify why
his reference classes are historically legitimate rather than ad hoc constructs
invented for Bayesian theatre. Historians do not infer non-existence from later
mythologisation, nor do they treat “mythic” literary development as
probabilistic evidence against an underlying person. Carrier’s priors therefore
predetermine the outcome, not because the numbers are extreme, but because the
conceptual machinery generating them already presupposes Jesus’ improbability
and Aristotle’s reality. Calling critics “emotional” is not a rebuttal – he
just hates the fact that I can see through his fatal methodological flaw.
Richard Carrier 7#:
“This is embarrassingly confused. Of course my article does give
“historically grounded” priors, entirely based on data that has passed peer
review multiple times now. It is not circular to observe the fact that “heavily
mythologized” persons usually don’t exist. We can count them and develop an
actual data-driven frequency of this; and that’s exactly what I did in my
original study, with a generous margin of error placing the prior odds on
heroes like Jesus existing over a whole third of the entire probability space.
The Gregor team’s study improved my results by narrowing that margin, yet
ending up within (and thus confirming) my original study’s tolerances, by
perfecting its empirical data. Their results now set the frequency of heavily
mythologized persons who plausibly existed at 1 in 4, even lower than I
assigned in my article on Aristotle, as I demonstrate in OPH.”
My comment: Carrier’s up to his old tricks
again because he equivocates between counting myths and establishing
historically relevant reference classes. The core error is not whether one can
“count” heavily mythologised figures, but whether the set he counts is
methodologically comparable to Jesus’ case in the first place. His dataset
conflates fundamentally different phenomena: purely literary gods, folkloric
heroes, euhemerised deities, composite figures, and later legendary accretions
onto real people, then treats them as a single frequency class. That move
already smuggles in the conclusion, because Jesus is classified as “heavily
mythologised” on the basis of the very sources whose historicity is in
question. This is textbook circularity: the evidence under evaluation is used
to determine the prior probability governing its own evaluation. No amount of
Carrier’s so-called “peer review” rescues that error. Moreover, Carrier
repeatedly asserts that mythologised founders “usually don’t exist,” but this
claim collapses once the comparison set is restricted to known historical
categories: named teachers, executed sect leaders, or founders of movements in
the Hellenistic–Roman world. In that properly constrained class, existence is
the norm, not the exception - even when extensive myth-making follows. And, I
mean, really – desperate appeals to this Gregor team does not save the
argument, as refining error margins within a flawed classification scheme does
not validate the scheme itself, and more than it does within the circle of
young earth creationists. Narrowing the confidence interval around an invalid
prior does not make it historically grounded; it merely makes the mistake more
precise.
I think that’ll do
for now. After you’ve read this, if you’re one of those still convinced by
Richard Carrier’s tricks, I encourage you to do what he’d absolutely hate you
to do; look behind the stage door; check not just his sources but the sources
of his sources, and the sources of those sources in context, and so on, and see
what they really amount to. See those sources in their full context, not in the
selectively trimmed form he presents. Ask whether the citations actually
support the claims he says they do, or whether they are being leaned on,
stretched, or quietly repurposed in typical Carrier style. Look closely at what
he calls “peer review,” which – like young earth creationism - frequently turns
out to be review by sympathetic presses or friendly associates rather than
sustained engagement by specialists in the relevant field. Notice how frequently
quotation replaces argument, how dissenting scholarship is collapsed into
caricature, and how methodological objections are waved away with appeals to
authority, metrics, or imagined consensus.
If you don’t want to
just take my word for it when it comes or his understanding of probability
theory and Bayesian methods, find someone you know who is an expert in the
field and ask them to evaluate both sets of arguments for you – I am immensely
confident they will tell you the same things I am about Carrier’s lack of
competence. The reality is, if you hold him to account rigorously and honestly,
the spell breaks - because the scaffolding holding up the Carrier structure
turns out to be far more fragile and undependable than he proclaims. The best
advice I can give you on Richard Carrier is to give him a wide berth.
