A few readers have indicated to me that my recent series of posts on the problematic use of Bayes Theorem for assessing “historical claims” have failed to make their intended point.
Hopefully here I can succinctly explain why Bayes cannot help us decide whether Christianity began with a historical Jesus.
Reason #1: If our question is simply, Did Jesus Exist? then it is meaningless. What is of interest is the question of how Christianity originated. What might Jesus have done that gave birth to the Christian religion? What did others do during the time of Jesus or after him that shaped or established Christianity? Those are the meaningful questions. Simply saying Jesus did or did not exist is somewhat pointless — unless, perhaps, one wants a negative answer in order to irritate believers.
Reason #2: If by using Bayes one concludes that Jesus “probably did not exist” then again, we have to ask, So what? If it appears unlikely that he existed then after weighing up the probabilities on the basis of the various strands of data, that tells the historian nothing useful at all. Simply saying that Jesus fits the pattern of mythical persons, if that’s where Bayesian inference leads, does not answer the question of whether he existed or not. Simply saying that there is, say, an 80% chance he did not exist still leaves open the possibility that he did exist. So what has been achieved? Nothing useful for the historian at all. Likewise, calculating that there is an 80% chance that he did exist would still leave open the possibility that he did not. The historian is no better off with either result.
I suspect King Philip II of Spain saw the odds of his Spanish Armada crushing the English fleet as overwhelmingly high. The odds against an event happening are irrelevant are irrelevant if they happen. And many times the unexpected and “out of the blue” does happen in history. That they may have been judged to have been unlikely at the time makes no difference to the fact that they happened and are part of the historical record.
Most historical events are “unlikely” or unforeseen until after they happen. After they happen commentators and the rest of us can see how “inevitable” they were. We can always predict what will happen after it happens. Carrier’s mythicist hypothesis can predict the type of evidence the historian will find after the hypothesis was originally formulated on the basis of that evidence. One might look at any number of events in the past and ask, What was the likelihood of X happening? The chances that I will be struck by lightning are very slim indeed. But if I were to be struck by lightning this weekend — stormy weather appears to be approaching — the odds against it happening will mean absolutely nothing against the fact (fingers crossed it won’t be a fact) that it “happened”! Odds against something happening are meaningless when investigating “what did happen”.
We don’t need Bayesian calculations to decide whether there was a Roman empire, or whether its emperors were worshiped as deities, or whether the Roman power destroyed the Jewish temple in 70 CE. The kind of evidence we have for the “raw facts” of the past, including who lived and who did what, are grounded in the same kinds of judgments we make in testing the authenticity of modern claims, whether they be events reported in the news or checking the reliability of advertised claims about a product that interests us. Some of us are less careful with respect to such matters than is healthy and easily believe false claims, present and past. When a historian is interested in whether “new facts” can be dug up to throw new light on a question, it is to the archives, to official records, to diaries and letters and reports of various kinds that they turn. These are tested for authenticity and reliability. If there is doubt about any detail it is more likely to find its way into publication by way of a footnote — with its questionable status clearly noted.
The only justifiable approach to reconstructing Christian origins is to build on the sources we have and on what we know about them — not on what we surmise about them. That approach will not allow us to join in the games of imagining what Jesus and his followers may have done. (We have stories of Jesus and we cannot assume — without justification that would pass the test in any other field of sound empirical inquiry — that they must be based on true events.) We will not have the wealth of details we would like if we avoid make-believe games. But the professional will not apologize for tailoring the question and scope of inquiry to accord with the extent and nature of the source material.
Sure, there is room for Bayesian probability when it comes to drawing certain kinds of inferences from archaeological data or for comparing the likelihood of competing hypotheses, but claiming that so-and-so did or did not exist is by itself a rather meaningless exercise for the reason I stated above.
(See also the section of my earlier post pointing out that not even postmodernist historians work with “what probably happened“.)
To address one specific point I referred to in my recent series: It may well be that one can find in literature more mythical persons who fit a Rank-Raglan hero type, but that is irrelevant to the fact that some historical persons did resurface in later literature wearing Rank-Raglan features (born of a virgin, died on a hill, etc). But even Raglan himself understood that the historicity of a figure was unrelated to the fact that fanciful tales were later told about him or her. If Jesus scores more highly than other historical figures on the R-R scale, so be it: such a “fact” would have no bearing whatever on whether or not he might have been historical. Ask Raglan himself.
Neil Godfrey
Latest posts by Neil Godfrey (see all)
- A Beginning of Christianity — An Objection; and the Role of Asceticism - 2025-02-21 01:00:10 GMT+0000
- A Beginning of Christianity? — A Closer Look in Antioch - 2025-02-20 11:35:52 GMT+0000
- Which One Came First? “Gnostic” ideas or “Orthodox” Christianity? - 2025-02-18 09:25:20 GMT+0000
If you enjoyed this post, please consider donating to Vridar. Thanks!
I am not a scholar, just a person raised and educated Roman Catholic who developed, in high school, a curiosity about the nuts and bolts of early church history. Research into what objectively is known and objectively can be known eventually led me to the following attitude: Given how deeply Jesus or the Jesus character is embedded in Christian-rooted cultures, I happen to find it immensely interesting to have concluded that we can’t know if Jesus or anyone like the Jesus character actually existed. As to the origins of Christianity, we assume that stories about Jesus or some sort of Jesus character were somehow at the root of what has come down to us as Christianity. The stories could have some degree of historicity or not — it makes no difference in their ability to inspire belief in or appreciation of a wide range of Christianities from fundamentalist to metaphorical or allegorical.
Hermann Detering was a Christian with a _principled_ objection to believing that Jesus was a historical figure. He seems to have felt that Christianity was too noble a belief system to be corrupted by reality (not a quote or even a careful paraphrase).
As another Catholic-raised present-atheist, that’s the kind of Christianity I could join.
Thomas Brodie was another. He likewise held fast to the Christian faith while believing the Jesus story to be a “noble myth” about God’s love.
Agreed. Yes, the question is of public interest and personal relevance to many. The most responsible answer a historian should offer is that we cannot know, that we have no evidence for his existence that is comparable to the kinds of evidence we have for other known historical persons.
The more fruitful historical question is this — if Christianity never existed, what would exist in its place?
For example, would Morality be much different than it is today? Bruno Bauer (the young Hegelian) showed how much New Testament authors owed to Roman Moral philosophers like Seneca the Younger.
John Dominic Crossan, for example, recently posited that a historical Jesus is hardly distinguishable from itinerant Cynic philosophers so common in the Roman Empire.
Even Hegel (1830) had shown Virgin Births, Divine Incarnations, Trinities, and promises of Immortality long before Christianity.
Even the Morality of persecution of gay people long predates Christianity — famous in ancient China and ancient Persia, for example.
What would the world look like WITHOUT Christianity? Nietzsche (“The Antichrist,” 1888) was mistaken, I say, to imagine a utopian world of Roman aristocrats worshipping Dionysus.
A quasi-Christian global civilization existed long before Christianity in Rome, Persia, and much of the ancient world. Monasteries and convents, for example, were unknown in Judaism and first century Christianity — but were centuries old in the Far East and already common in Rome.
As for a historical Jesus — the scholar Morton Smith (not an apologist) did much to recognize the flesh-and-blood figure of Jesus in his book, “Jesus the Magician,” (1977). It does not matter that Magic is superstition — what truly matters is that Magicians were always a historical typus — and everything about Jesus matches that typus.
What made Jesus of Nazareth unique was that he was the first (and perhaps only) Jewish prophet who was also an exorcist — and the first (and perhaps only) exorcist who was also a Jewish prophet. Yet both are recognizable social types from the ancient Levant.
As for the Jewish doctrine of an End Time, a General Resurrection, and a Last Judgment in which the righteous are sent to heaven for an afterlife with angels, while the wicked are sent to hell for an afterlife with demons — all that is much older than the Jewish prophets.
Apocalyptic culture reaches back to the Gathas of Zarathustra, which date before Exilic times (~540 BC; perhaps earlier than ~900 BC, according to some historians).
It is no coincidence that Apocalyptic religion began with the Persian “Farsi” priests and appeared in Judea among “Pharisee” priests only after the Persians had populated Yehud (~540 BC) in Judea.
Nor is it a coincidence that the so-called Apostle Paul was raised as a Pharisee before becoming the most influential of all Christians. Nor is it a coincidence that whole segments of Paul’s Epistles match whole segments of the writings of his contemporary, Seneca the Younger — as demonstrated by Bruno Bauer (1877).
If not Christianity, what? That’s the question historians should ask, in my reading.
No that question isn’t useful for historians either. Counterfactuals aren’t particularly productive in historical analysis because there is simply no way to predict or determine what would have been there instead of Christianity. As Neil points out, you can’t account for so much of history, because so many unlikely and unpredictable things happen. So, these sort of “what if Christianity never existed?” questions are completely pointless except to science fiction and historical fiction authors who want to sell a novel or two. Actual historians shouldn’t waste time on useless hypotheticals that will accomplish literally nothing.
It’s like the philosophical trolley problem (which has been criticized as a semi-useless tool, and non-educational for decades now). A complete waste of time, with no real application or relation to morals in the world, and only exists for college classrooms, so students can meander over a ten page term paper. These kinds of thought experiments are, functionally, useless except as entertainment. Same goes for these counterfactual “what-ifs” of “what if Christianity didn’t exist?” So what? Who cares? It doesn’t change the world we live in, even if you could reasonably answer that question (which you couldn’t).
Re “What might Jesus have done that gave birth to the Christian religion?”
This implies that Jesus was the founder of Christianity and it seems not to be the case.
Point 1: Is Christianity more about Jesus or about Paul?
Point 2: In the book “The Jesus Hoax: How St. Paul’s Cabal Fooled the World for Two Thousand Years” the author, David Skrbina makes the case that Paul created Christianity to undermine the Roman empire and save Judaism. This not outlandish and fits all of the data and also provides a motivation for the spread of Christianity.
Point 3: In the book “After Jesus Before Christianity: A Historical Exploration of the First Two Centuries of Jesus Movements” it is made clear that “Christianity did not exist in that time period. There were so many types of groups one could categorize as Jesus followers, that one has to ask. If the people of that time couldn’t figure out what Christianity was, why would we think we can?
With regard to your point of using Bayes theorem on historical issues (and my experience is almost exclusively from reading Richard Carrier’s works, the objection I had was that Carrier was too generous in providing priors to many of the questions he addressed. Since even with those oh, so generous priors assumed, the probabilities of there being many truths available was still minimal, I guess that is why Carrier did his analyses that way.
And, as far as saying Bayes theorem is not applicable, that is nonsensical. It is a tool, designed to evaluate claims, and using it is no different from using a magnifying glass to see something small.
Re “Simply saying that there is, say, an 80% chance he did not exist still leaves open the possibility that he did exist. So what has been achieved? Nothing useful for the historian at all. Likewise, calculating that there is an 80% chance that he did exist would still leave open the possibility that he did not. The historian is no better off with either result.”
So, his doesn’t even serve as an assessment of the current state of the evidence? You aren’t looking for proof, are you? All historical facts are probabilistic in nature. There is no such thing as proof.
Re “Sure, there is room for Bayesian probability when it comes to drawing certain kinds of inferences from archaeological data or for comparing the likelihood of competing hypotheses, but claiming that so-and-so did or did not exist is by itself a rather meaningless exercise for the reason I stated above.”
So, it is not the use of Bayes’ theorem that is problematical, it is the questions addressed by the researchers? So, why is the title of this piece “The Folly of Bayesian Probability in Doing History”?
This is the first piece of yours I have read, and I have been reading them for a long time, that seems strangely biased.
“So, it is not the use of Bayes’ theorem that is problematical, it is the questions addressed by the researchers? ”
The trouble is that probably all interesting historical questions are too complex for statistical analysis. Mathematics in general, including its half-brother statistics, deal only with idealized and exsanguinated abstractions, and are not applicable to the real world.
That said, Carrier’s book has a lot of interesting content; but it was silly to try to precisely quantify his analysis.
I disagree. There is nothing probabilistic about “the fact” that Japanese aircraft bombed Pearl Harbor in December 1941. There is nothing probabilistic about the fact that Rome once ruled the Mediterranean or that her armies destroyed the Jerusalem temple in 70 CE. Those are known facts empirically established.
I know one hears so often that we can say nothing certain about the past but that claim is confusing two different things. I illustrated these two different things with the way Japan’s attack on Pearl Harbor can be described by historians. The fact of the event is not a probability. What is open to debate are other questions arising from that known and sure fact of history. See my illustration of this point.
Nothing wrong with Bayes as a tool for calculating probabilities. Nothing at all. But what I disagree with is that “doing history” is fundamentally about “probabilities”. Historians study the past events. Those events are on the whole known facts empirically established. They are not, for most part, probability statements.
Historians explore questions arising from those empirically established known and certain events of the past.
I think the fundamental point of difference between your view, Neil, and those who disagree is summed up in these quotes from you:
“Historical events are unique and not subject to probability calculations.”
“The fact of the event is not a probability.”
“Those events are on the whole known facts empirically established.”
Those of us who disagree with you, I think, disagree here. Petr spend a long time explaining the maths, partly I think because the probability does not get to 1. It just doesn’t, that’s not how it works! It also shifts more than you might expect after a few iterations.
Similarly, ALL reasonable claims came to be credible through at least subconscious Bayesian reasoning. So the (maybe subconscious) reason you don’t doubt the Pearl harbour narrative is because you have gone through a process of assessing multiple bits of evidence **in exactly the way RC does for the Jesus example** and the result is very close to, but not actually 1. When people reason validly, that’s what they do. I imagine you disagree with this assertion!
Also to quick point on the lack of multiple ‘rolls of the dice’: it is not the event itself that is repeated, but bits of evidence considered one after the other. Exactly as has been done to establish the Roman Empire and WW1 and Pearl Harbour. There is mountains of ‘expected’ evidence which would be ‘predicted’ on the hypothesis that WW1 happened. All this evidence can be plugged into a Bayesian calculation. There is not some magic, separate process for determining these ‘known’ facts that differs from false or unsure ones.
Is there not a difference between empirically determining that an event happened or a person existed and assessing the probability that an event or person happened? The difference is one of method and conclusion.
Carrier speaks of certain types of information being beyond Bayes — such as knowing for a fact that you are reading these words on a screen right now. But Carrier would have done well to have paused to think through that exception and to ask where the dividing line might be between empirical and probability determinations. He would have had to conclude, I believe, that there is no sliding scale but a difference in types of knowledge and the grounds for different types of “knowing” — that is, two different types of knowledge each in their own buckets. Not a sliding scale.
Our belief that we really are reading these words right now is empirical knowledge, not probability assessments. We are relying on the evidence we can see before us to know, empirically, that we are reading these words. Empirical knowledge extends to what we learn from reading inscriptions on coins or monuments uncovered by archaeologists. There is room for debate over the contents of what we read, but that someone produced those coins and inscriptions is empirically determined and not a question of probability assessment.
Our knowledge or belief that Japan attacked Pearl Harbor does “reach” 1 so there is no room for any doubt at all. The only room for doubt I can see here is if we go all the way and say we cannot know anything, period, such as not even knowing for sure that we are reading these words right now. In that case, literally everything is probable so that probability itself loses its meaning and usefulness as a concept.
We can choose in retrospect to apply Bayesian reasoning to whether the Pearl Harbor attack happened, but that is a theoretical exercise in retrospect and is not the reason we know it happened.
It is the difference between empirically derived information and probability assessments.
But in history we do know that certain events happened in the past and certain actors did such and such, etc. There are no doubts with empirically established information. That Julius Caesar existed and did certain things (conquest of Gaul, reforms in Rome, etc) has no room for probability assessments except in retrospect as a theoretical exercise. We have the independently confirmed and established evidence that leaves no room for doubts as to the basics. The historian does not dispute those. Questions arise over motives, consequences, details but not the core events.
In another discussion thread (Part 2), I have expressed my (vague) future plan to write a summary of my understanding of what is going on here.
Now I just feel like adding a remark on my impression that the basic notions around which the discussion evolves are not clarified; then the discussion can be endless, making not much sense after all.
Instead of Pearl Harbor, let me take the Apollo moon-landing event (that I saw on TV as a boy) as an example, which can hopefully lead towards a clarification of one basic notion.
I voice several assumptions, testing if we all agree on them:
1/ One can create a scenario S in which the Apollo moon-landing event in 1969 did not happen, since it was only faked; we know that such scenarios were indeed suggested.
2/ If such a scenario S is very carefully elaborated then all the evidence that we have cannot really contradict S.
3/ The meaning of the word “impossible” (for the use in this discussion) does not allow us to say “it is impossible that S of the case 2 is true”. We are only justified to say “S is extremely improbable”.
4/ If a scenario S is not impossible but “only” extremely improbable, then the probability that S is true is non-zero (for us, who only have the evidence that we have). This entails that the probability of any scenario S’ that contradicts S is not 1 (for us).
Exactly so — your shift in an illustration for the point of discussion only illustrates my point. Conspiracy theorists jettison empirical testing of ideas and substitute question-begging, confirmation-bias notions. Of course Bayes is great for challenging conspiracy theories. A perfect and valid use of Bayes.
But you cannot apply probabilistic reasoning to the empirically established events of history for reasons I have attempted to explain.
Petr, a constant theme in your responses is, it seems to me, an attempt to explain the process of Bayesian reasoning as new information comes to light. I understand the process and the mathematics used. See, for example, my use of Bayes theorem applied to the question of the “Brother of the Lord” — https://vridar.org/2012/04/22/putting-james-the-brother-of-the-lord-to-a-bayesian-test/
I earlier posted a link to around 50 other posts of mine about Bayes. To keep it simple, here is just one more: https://vridar.org/2011/06/08/bayes-theorem-and-the-jesus-mythicism-historicity-conflict/
I believe I do understand the process. I “believe in Bayes”, you might say — but I do not believe it can be used for one off contingent events that are empirically established. And every known event in history falls into that category. We know Julius Caesar was assassinated by empirical confirmation from the sources. Probabilistic reasoning used to confirm that event and it is not an event that has any smidgen of a degree less than a probability of 1.
My sentence “I voice several assumptions, testing if we all agree on them:” was meant as a yes/no question; let me formulate the question explicitly:
Neil, do you agree with my assumptions 1,2,3, and 4? Yes, or no?
Why do you ask? What are you trying to argue? I agree and acknowledge that Bayes theorem is a valid tool in many situations. Are you trying to convince me to accept what I already accept? Do you understand what my point is? Can you encapsulate in your own words what I am arguing?
Petr — if you are serious about an honest and serious discussion, I implore you, please try, just try, to set out what you believe my argument to be. What is it you think you are trying to argue against? Please … I am really confused by what seems to me to be a wild series of attempts to simply avoid the point I am trying to make. I have asked for this more than once, now, and you seem to refuse point blank. If you continue in your refusal I will begin to think you are a troll.
But if you think it is relevant, here are my responses to each point:
We can imagine anything we like. What is your point and how does it relate to my argument? You asked if I agree with your “assumptions” — I don’t see any “assumption” here, I only see a point blank proposition. What is your point?
So? What does this have to do with empirically establishing an event in reality? Are you really positing a scenario where I have to accept only evidence you allow for a particular mind-game? Αnything can be proved if we make up rules like that. What is your point? Again, you are not presenting an assumption here but rather a point blank proposition.
I don’t know of any scenario that I have encountered arguing for the moon landing being a hoax that leaves anything that is not “contradictable”. Are you trying to tell me that there are water-tight arguments that the moon landing was a hoax and no-one can dispute those arguments? If so, I have to disagree with you.
But this is becoming tiresome. I am beginning to think you have not bothered a moment to try to understand my argument and are only flying off the handle in some misguided attempt to say Bayes is a valid way of reasoning about everything.
Again, what is the assumption you are asking me to agree to? Of course if you set the rules as to what is impossible etc you leave me no choice but to play your game. Now please Petr — I have attempted to understand your point of view so please try to understand mine. What is it that you are arguing against? What is it that you think I believe that you have to prove wrong?
>4/ If a scenario S is not impossible but “only” extremely improbable, then the probability that S is true is non-zero (for us, who only have the evidence that we have). This entails that the probability of any scenario S’ that contradicts S is not 1 (for us).
So what? Do you not understand a word of my argument? Do you know the difference between empirically established event and a probability of an event?
Are you trying to prove to me that the Pearl Harbor bombing “probably happened”? It is your turn to answer my question. Is that what you are saying?
Are you trying to persuade me to agree that there is no certainty about ANYTHING (that is, “anything” beyond something as basic as ‘knowing you are reading these words right now’) in life and that EVERYTHING is a matter of degrees of probability? Is that your position? If so, please say so directly.
What do you understand by empirically established fact? Does such a concept exist in your thinking?
If you are not simply a troll then answer my questions.
Neil, I have no interest to engage in an emotional discussion.
I felt indebted to you, since I read many interesting texts on your blog in the past. Now I was hoping to provide some useful feedback to you, but this obviously has not worked. So I am now just sending best wishes to you.
(Of course, you can have a last word here if you wish, it is your blog anyway; but I do not plan to react anymore.)
I simply asked you to tell me what it is that you think I misunderstand or what, specifically, I am in error in stating. Asking for you to clarify your disagreement with me is not “an emotional discussion” from my part. But your refusal to do so despite my repeated requests does wear my patience a little. I believe I have been very patient and “unemotional” with your responses, taking time to respond fully. I only ask you do the same with my requests of you.
Proposing fanciful scenarios in an effort to respond to my real-world ones does lead me to think you are not grasping my point. I am disappointed you are bowing out without explanation or clarification of what, exactly, I have misunderstood about Carrier’s use of Bayes — after all the time I have taken to try to engage with you.
Perhaps someone else who has been following this discussion might like to take Petr’s side and try to explain to me what exactly I do not appear to have understood about Carrier’s use of Bayes.
I re-read my earlier reply and can see how you took it as “emotional” — but I can assure you I was quite serious and and honest in my questions — they were sincere questions, not emotional reactions. I really am trying to understand what it is that you believe I misunderstand. I am getting quite lost with your alternative scenarios of which I am having great difficulty seeing their relevance to my position.
I am still once more responding, so that you do not think that I am impolite or that I am unable to “swallow” your sentence “If you are not simply a troll then answer my questions.”
(Btw, I am aware that there is a -nonzero- probability that you wrote, in fact, “If you are not simply tired …” but on the way from your keyboard to the server the sequence “tired” changed to “a troll” by errors of the communication lines. Nevertheless, I use Bayesian reasoning, and though my prior expectation (also called probability) of your possible using of “troll” was low, now, by getting the evidence I have, my posterior confidence (probability) that it happened is virtually 1.)
In my opinion, the main problem of your several-part treatise can be demonstrated on your currently last sentence “Empirically established knowledge is not the same as Bayesian probabilistic knowledge.”, by which, from my viewpoint, you create a false dilemma.
I suppose we both agree that no historian is born with the knowledge whether or not Japan attacked Pearl Harbor, or whether or not Pontius Pilate ordered Jesus of Nazareth to be crucified, etc. etc.
But you seem not to agree that such events, as well as surrounding circumstances etc. etc., should be approached (by the historian) with -the same kind of reasoning-, which can be characterized as Bayesian reasoning.
Neil, you might surely conclude that I (and some other people, including Carrier) have not understood your point, though you have written so much on this here.
But do not take it as impolite, please, that I really can see no sense in continuing this discussion, so I -really do not plan to react anymore-.
A troll is someone who ignores what is being said and continues their own agenda without engaging with the logic of the other. I have implored with you repeatedly to identify where you see me to have misunderstood Carrier only to be given reasons you cannot do that, from what I have understood of your comments. That, and the ever more bizarre fantastical examples being proposed in preference to empirically established events. That was what led me to begin to wonder if you were trolling. I regret making that comment and apologize. I should have been more patient.
Well if you won’t reply any further I will leave you with the final word of argument. (I did post a reply a while ago but decided to delete it, thinking it added nothing to what has already been said.)
Carrier has only responded to my posts, as far as I am aware, with the false rebuttal that I was claiming he was saying his theory predicted something very specific about the sources. The only criticism of his that I have seen has not attempted to address the point I was making.
As I wrote (in the last year 🙂 ), I did no plan to react anymore, since I felt no sense in a further discussion. Now I have got a new impulse when seeing that Richard Carrier has again reacted, and you have written “I think we are talking past each other.” to him (similarly as to me previously).
This also reminded me of Tim W.’s sentence (in another context) “The political left, center, and right sometimes talk past one another because they often use the same words to mean different concepts.”
I also recall my sentence “Now I just feel like adding a remark on my impression that the basic notions around which the discussion evolves are not clarified; then the discussion can be endless, making not much sense after all.”
I hope that we agree that it is a necessary prerequisite of any reasonable discussion that the meaning of the used notions is clarified among the discussing parties. So I try to do one more attempt to clarify at least one issue.
The word “probability” is generally used in various contexts. One context is exemplified by random processes studied in pure mathematics, another context is the case of criminalists who try to find out what has most probably happened ….
One axiom that is generally accepted can be formulated as follows:
if a hypothesis (aiming to explain a past event, to be more concrete) cannot be shown to be impossible by the evidence that we have, then it must have a nonzero probability for us (which might be extremely small but is necessarily greater than zero).
Does the concept of probability that you have in mind satisfy this axiom?
(I hope it is clear that it is a YES/NO question.)
Again I think we are talking past each other 😉 .. I have always tried to point out that I fully agree that Bayes is just fine for assessing hypotheses about what might have happened. Richard appears not to have responded to the content of my posts but only to (“decontextualized”) comments that were part of a subsequent exchange. What I would appreciate is a response to the core arguments of my posts. I get the impression — you can correct me if I am mis-stating things here — that I am thought not to understand Bayes and how it applies to hypotheses. But that is not what my posts are about.
(Technical remark. I have been trying to post this several times but it seemed to fail.)
You write “Again I think we are talking past each other”; to overcome this, I try hard to clarify how you understand the notions that you are using.
But instead of getting a clear answer to my question, I get
“What I would appreciate is a response to the core arguments of my posts. I get the impression that I am thought not to understand Bayes and how it applies to hypotheses. But that is not what my posts are about.”
You can see that it is unclear to (at least some of) the readers what your core arguments really are.
First of all, you can only be understood if the readers understand the concepts that you have in mind (and that you have not really explained in your long texts that touch on many issues).
Your former comments have already demonstrated to me that you have not thought through the core of the Thomas Bayes’ ball experiment that you yourself recalled.
(E.g., you said “… after being allowed only three guesses. He would say, no, that’s not what
this theory is about.” [I suppose that by “guesses” you meant the results of the described experiments.] I tried to explain that in such case TB would simply calculate an interval inside which the ball is with probability 95%, say. If he gets more evidence, then the calculated interval gets narrower; that’s all. This has nothing to do with needing to have countless experiments at our disposal, or so.)
Also, you have not recognized a correct application of Bayes’ theorem, when writing
“… your calculations are erroneous within the constraints of a limited number of throws. (Endless throws is something quite different and exactly where probability theory works – with data that is theoretically infinite.)”
It is frustrating for me that you neither demonstrated the errors you claimed I was doing, nor withdrew your claim. Therefore I am afraid that you still do not understand that I demonstrated a standard usage of Bayes’ Theorem … 🙁
I could return to this in more detail, if you are really interested, but here I would like primarily to clarify your opposition to Carrier’s axiom
Axiom 4: Every claim has a nonzero probability of being true or false (unless its being true or false is logically impossible). . .
Please, let us restrict ourselves just to the claims on what and how something happened in history,
which is the crucial subject here. You seem to think that you have demonstrated that it is reasonable to not accept this axiom, though you have not demonstrated any instance where this axiom can lead (or even has lead) to some unreasonable consequences.
Your examples like “Japan attacked Pearl Harbor” are surely not such demonstrations.
(I hope that I do not need to stress that I cannot go and observe that Japan attacked Pearl Harbor. I can only study the evidence that I have at my disposal now, and conclude that I cannot -reasonably- doubt that it happened; the hypothesis that it did not happen, while we still have the evidence that we have, is extremely improbable; that’s all.)
You seem to simply insist that we must not say “extremely improbable” but we must say “impossible, with probability being the absolute zero”. This is one of your core arguments?
I feel that you might say that your opposition to Axiom 4 does not belong to your core arguments after all. If this is the case, then you should clearly say this and hopefully withdraw your opposition to Axiom 4 altogether (unless you clearly demonstrate where Richard Carrier was mislead by using this axiom).
When this issue gets clarified, we can proceed further if you wish.
I don’t know why but your comments were going straight to the blog’s spam folder. Hopefully that will not happen again (I could see no reason for the spam filter being activated — neither your name nor email or IP address were listed there.)
I am trying to zero in on the difference between “the facts” historians work with and hypotheses they construct to explain those “facts” or “events”. The events themselves are not hypothetical statements. They are empirically established. There is no probability about them except the “probability” of 1, which is really not a probability ratio estimate at all.
Carrier confuses two things when he refers to “claims” by historians. From my reading of Proving History (and I have tried to support my interpretation of what Carrier has said with relevant quotations) he treats as common “claims” both empirically established events and hypothetical explanations for events. This, I believe, is an error. Only the latter are open to Bayesian assessments. The former are not established on the basis of probabilities but on the basis of hard and undisputed evidence that is agreed can have no other explanation. Historians did not examine the probabilities of this and that in order to weigh up if it was likely that Hitler came to power in Germany in 1933. There is a difference between empirical facts and probability hypotheses.
Yes, it is possible for us to sit back and look at all the empirical evidence for Hitler’s assumption of power and weigh up probability estimates for each bit of it, but that’s all an arm-chair exercise in retrospect. It is not how the event itself was known to have happened. Carrier might say that subconsciously we are in the back of our minds weighing up frequencies of this and that, but as I try to point out in my latest post that is not how historians think.
This is part 2 of my reply to your above comment:
I read your disagreements with me, but I thought at one point — I may be wrong — that you said you did not understand my point of view in a way that you could explain it in some kind of ah-ha moment. I may have misunderstood you.
If my texts are too long to engage with or understand, I guess they are not for everybody. If parts are unclear people can ask for clarifications.
Yes. Pearl Harbor is only an example, but the point is Yes. Some facts — Carrier refers to the fact of reading these words right now — are clear facts with no probabilities about them at all. Yes or No. Black or White. What he is saying is that direct empirical evidence is clear cut and no probability or Bayesian analysis is needed. I agree. But I believe he misunderstands the nature of the evidence or reasons we believe certain historical events happened in the past. I believe he is confusing hypotheses and explanations with our empirical knowledge the actual events themselves.
I may have misunderstood or not been as aware of Carrier’s position as I should be, but I am left with the impression, right or wrong, that he has not engaged as fully with the question of the nature of historical knowledge (as discussed among historians and philosophers of history) as he might.
Neil, I suppose you would surely feel unfair to be viewed as “someone who ignores what is being said and continues their own agenda without engaging with the logic of the other”. (I am using your own words here.)
But when you look back at your reactions to all critical remarks, can you at least somehow understand that some of us can be getting a sort of such an impression?
(I sincerely hope that you do not mind not taking yourself too seriously, and that you thus take my comment with a sense of humor.)
Here I try to recall two concrete issues for your possible reconsideration, one mine and one Geoff’s from the top of this thread.
1/ I could not believe that you make no conceptual difference between “impossible” and “extremely improbable” when we try to describe what happened in the past; I hoped to clarify the situation by presenting my extraterrestrial and other examples.
Now I understand that you indeed make no difference between such concepts. It is your choice, and it is surely reasonable from the practical point of view. (Even our justice systems, when sending people to prison etc., are satisfied with proving the convicts guilty beyond a reasonable doubt, not necessarily beyond any doubt.)
But I find your criticism of the (cautious) axiom “Every claim (on what happened in the past) has a nonzero probability of being true or false unless its being true or false is logically impossible” as completely unfounded. You have demonstrated no single concrete instance when the application of such an axiom has led Carrier (or another scholar) to an unreasonable conclusion; you were simply inventing strawman-examples.
So, please, either show an actual example where this axiom has clearly misled someone, or admit that your criticism is unfounded.
2/ Look at Geoff’s comment above (Geoff 2024-12-14 11:10:07 GMT+0000 at 11:10).
In my viewpoint, your response “Is there not a difference between empirically determining that an event happened or a person existed and assessing the probability that an event or person happened?” completely misses the point Geoff is making.
Can you reconsider this, please?
I have set the agenda with my posts. I have been at a loss to understand how your comments relate to my arguments. (Looking back I can see why you were rightly offended at the tone of my replies and I apologize again on that score.)
I do not say Carrier reaches “unreasonable conclusions”. That’s why I haven’t supplied evidence that he has done so. 😉
Carrier writes “Axiom 4: Every claim has a nonzero probability of being true or false (unless its being true or false is logically impossible). . . . All claims have a nonzero epistemic probability of being true, no matter how absurd they may be (unless they’re logically impossible or unintelligible), because we can always be wrong about anything. And that entails there is always a nonzero probability that we are wrong, no matter how small that probability is. And therefore there is always a converse of that probability, which is the probability that we are right (or would be right) to believe that claim. This holds even for many claims that are supposedly certain, such as the conclusions of logical or mathematical proofs.”
I disagree with Carrier’s words above. I cannot agree that a historical claim that Japan bombed Pearl Harbor is even possibly — however remotely — wrong. I don’t see that as a straw man example. It’s a test case of Carrier’s words. Carrier is asking us to accept that however small the probability, we could be wrong about that claim just as surely as we could, however unlikely it might seem, be wrong about the calculation of pi.
I don’t accept that.
What Carrier is doing is trying to force probabilistic reasoning into every situation of what people claim to know — except for (his examples) “immediate experiences that at their most basic level …. that you see words in front of you at this very moment, or that ‘Caesar was immortal and Brutus killed him’ is logically impossible”.
I disagree with that attempt. Most historians that I have read who discuss their methods disagree — including those listed by Carrier. (I am thinking of doing a post placing quotes from those historians Carrier cites against his own explanation of historical reasoning. Carrier wrongly asserts their methods are at bottom Bayesian without them realizing it.)
The one person the axiom has misled is the one who proposed the axiom in the first place — Richard Carrier. All historical claims are not determined by probabilistic reasoning. Far from it. Carrier argues that they should be and that historians are deep down, “beneath the lid”, really appealing to probabilities, but by his explanation then if probabilistic reasoning applies to pretty much everything then it is meaningless as a distinguishing and defining term.
I gave an example in my latest post on Carrier’s view: the reasoning that a historian applies to suspecting a murder plot is background knowledge of what is required to take full control of a state treasury. That is the background knowledge the historian also appeals to in the reader. I cannot believe, and there is no evidence to support Carrier’s assertion, that the historian somehow was perhaps subconsciously letting known instances and frequencies of other takeovers of bureaucracies guide the historian’s conclusion. Such an explanation is far less probable (in Bayesian terms! 😉 ) than the former explanation — that the historian was drawing on background knowledge and not probabilities or frequencies.
If it appears that I have avoided your questions then I apologize. But I find some of your questions going past me because I cannot understand how they relate to what I’ve attempted to explain. I will promise to be more patient and watch my tone in future.
Now to address your other request re Geoff’s comment….
As I understand Geoff, he is saying that we can look at all the evidence and apply Bayesian reasoning to it all and come up with the historical events the evidence points us to. Yes, and I think I have said this before, that we can do an arm-chair reflection of all the evidence and do that theoretical exercise. But all of that is in retrospect. It is not how historians came to know the events happened in the first place.
Carrier says that “beneath the hood” if we look we can see the reasoning was based on probabilities. And Geoff argues that we can examine the evidence in Bayesian terms and come to the conclusion that historians have reached about what happened. But that’s all in hindsight and it does not prove that that’s how anyone came to know those events happened. Simply saying that a Bayesian look can lead to result X, therefore the historians used — at some deep unconscious level perhaps — saying that therefore historians used Bayesian reasoning does not follow.
Take for example the evidence for World War 1. Geoff writes:
This is circular reasoning, not Bayesian reasoning, in my view. In hindsight it is easy to say that this and that piece of evidence “predicted” WW1 — but it doesn’t really. The evidence has been selected because it meets the criteria of being able to predict WW1. If historians could look at evidence and make genuine predictions from it they could do the same with the evidence around us in the world today and predict the future.
Can I ask what, exactly, you see me as arguing that is in error? What, exactly, have I misunderstood about Carrier’s writings?
A few points to the above three reactions of yours:
1/ No need to apologize again for the former tone of your replies. I surely appreciate your overall friendly approach and what I find as your honest desire for mutual understanding
(that’s why I am still reacting). My comment tried to draw attention to the fact that it seems that none of the critical remarks of all readers here has lead to a slightest change of your position so far. This is simply surprising for me (I have not expected that 🙂 ).
2/ You say “The one person the axiom has misled is the one who proposed the axiom in the first place — Richard Carrier.”
You say that you do not accept this axiom. I can say that I personally have no problem to accept this. All our knowledge about the past is clearly a function of the evidence we currently have, and I do not see any reason to insist that we must single out some past events for which we must claim -absolute- certainty.
Btw, if I tell you that by such singling out you would at least subconsciously use Bayesian reasoning anyway, then you will protest, since you do not do this by sitting down in the armchair and starting some probability calculations …
I find such protests of yours simply as terminology misunderstandings, and I thus cannot see that you have exposed any real weakness in the respective reasoning by Carrier and others.
3/ Regarding the reaction to Geoff, I am afraid that you keep missing the point. You say
“In hindsight it is easy to say that this and that piece of evidence “predicted” WW1 …”
though Geoff wrote
“There is mountains of ‘expected’ evidence which would be ‘predicted’ on the hypothesis that WW1 happened.”
You see, nobody was talking about predicting WW1, so you just confirmed my previous suspicion that you mix different uses of the notion of probability in your texts:
It is quite a different thing to try to assess the probability of future “contingent” events on one hand, and to try to assess our confidence in understanding what happened in the past, based on the evidence at our disposal, on the other hand.
It seems to me that you are just misled by the use of the word “predicts” when we speak on assessing the hypotheses concerning the past.
I also find your claims like “This is circular reasoning” absolutely baseless.
All this is just a terminology mismatch, in my opinion.
The reason I do not accept the claim is because Carrier confuses different terms under the one label “claim”. I tried to point out somewhere that Carrier brackets together as “claims” both the “factual events” themselves ALONG WITH interpretations and explanations and hypotheses about those events. Not all “events” have a “nonzero probability of being false”. That is simply not so.
Carrier frequently confuses different things. He fails to provide clear definitions of his terms at the outset and this is where much of the confusion of his argument arises.
I do not believe in mind-reading as a valid criticism. We need to look at the evidence that explains or otherwise a particular “claim” in the mind of a historian. There are far simpler reasons for much of our knowledge. To the one with a hammer, everything looks like a nail. There really are other tools in the historian’s toolkit that Carrier simply refused to recognize.
You are quite correct and I did once again go on autopilot with that comment.
But the point you are making is the other point I have been making in the posts and in these comments: There can be no such thing as a “hypothesis that WW1 happened”. That is not a hypothesis. This IS where circularity enters. Even if we could bizarrely say “I have a hypothesis that WW1 happened”, then one has to ask the basis for that hypothesis. It can’t come from thin air without any awareness of history. The hypothesis (if we call it that), arises BECAUSE of the evidence in the first place.
This goes back to my point that the Christ Myth theory exists BECAUSE certain types of evidence exist. Carrier then claims the hypothesis “predicts” or “expects” certain types of evidence.
That is invalid circularity.
I’m not sure you understand what my position is — 😉 (Can I ask again: what is it that you believe I misunderstand about Carrier’s position?)
For me it is interesting that you say
“Carrier frequently confuses different things. He fails to provide clear definitions of his terms at the outset and this is where much of the confusion of his argument arises.”
while I have such an impression when “Carrier” is replaced with “Neil Godfrey”.
I hope, Neil, that you are not offended by my frankness; I am just reacting to your general bold claim that only makes sense if it is carefully demonstrated by concrete (non-strawman) examples.
Regarding Carrier, I accept that he could/should have been more careful with his formulations using the words like “prediction”, to avoid any possible confusion. But in his first reaction to your current texts he wrote:
“This analysis only requires a single sentence to refute: the word “predict” simply means “makes probable,” and a prediction becomes “evidence for” a theory when that theory makes the predicted outcome probable and no other (plausible) theory makes it as probable (i.e. when the other theory does not predict that outcome, or not well).”
I can say that I understood this well when reading his book OHJ, but you seem to have not understood this (so far).
So you can take this as a partial answer of mine to your question “what is it that you believe I misunderstand about Carrier’s position?”
Two concrete points regarding you:
Your formulation “Not all “events” have a “nonzero probability of being false”. That is simply not so.”
is not quite clear to me but I guess that you meant it as “There exist claims on what happened in the past whose validity is -absolutely- certain”, and you find insisting on the word “absolutely” important, though you have not demonstrated how rational persons can be misled if they do not insist on this “absolutely”.
So here you do not seem to misunderstand Carrier, but your criticism of his axiom seems vacuous to me.
Your claim “That is invalid circularity.” was dealt with by several people here but you seem to simply repeat your claim on circularity without engaging with the logic of the responses.
Finally, I can say that I have no ambition to somehow argue with you. From the very beginning I was hoping to provide some feedback that you might find useful.
If you do not find my inputs anyhow helpful, then we should really stop our discussion.
What would be helpful and positive is if you would give specific examples of what concepts I have confused.
There was no confusion with Carrier’s use of “predict”. It is you who assumed — wrongly as I tried to explain — that I was confusing the meaning of his word “predict” and that “expect” would have been clearer to the mark. But I fully understood from the beginning that that’s what Carrier meant. It makes no difference to my argument. I tried to explain that I understood that his use of “predict” embraced the meaning “expect”.
That was a useless troll-level comment. Carrier was not prepared, it seems, to bother to try to demonstrate his claim with even a single point I made in the post. It was a cheap and lazy response.
So you say. I disagree. When I say something that Carrier said you say he didn’t say or mean that — so I point to where I have quoted him. What, exactly, do you think I have misunderstood about Carrier’s argument in Proving History? …. Okay — I see you have answered that question as I read on….
Let me try to explain again where the confusion lies here. I was referring to Carrier’s Axiom 4. Here are Carrier’s words which I was addressing:
Axiom 4: Every claim has a nonzero probability of being true or false (unless its being true or false is logically impossible).
If you read Carrier’s Proving History it will readily become apparent that by “claim” here Carrier includes both “events” as well as “hypotheses” about events. That is where his conceptual confusion begins — lumping events with hypotheses about events under the catch all term “claims”.
The reason I disagree with Carrier’s axiom 4 is that I do not believe “events” per se have a nonzero possibility of being false. Yes, hypotheses have a nonzero possibility of being false. But not all events do. Does anyone really believe there is a nonzero possibility that there was no Second World War? Or First World War? Or American Revolution/War of Independence? Of course not. Carrier’s sweeping claim is simply not so.
No. The criticism of mine that you cite (invalid circularity) was NOT a criticism of Axiom 4. It was a criticism of another claim — the one you said I had not given an adequate response to earlier. The “circularity” criticism was directed at the claim that a “hypothesis of WW1 predicts/expects certain kinds of evidence”. Carrier has the same circularity when he says the Christ myth hypothesis “predicts/expects” certain kinds of evidence. The reason it is circular is because the hypothesis itself was born from the same “certain kinds of evidence” that it then turns around and “predicts/expects”!)
Addition:
ChatGPT told me
“The word “vacuous” can carry a negative connotation, as it implies that something is empty, lacking in substance, or thoughtless, which might offend the person being addressed if not delivered carefully.”
and suggested that I write
“I find your criticism of this axiom unconvincing.”
Ok, I thus apologize and follow the recommendation 🙂
No need. The context sets the tone, rarely a single word. I want understanding between us.
By the way, with respect to something you said in your other comment, it never occurred to me that we were arguing — but rather discussing and trying to understand each other.
So I understand that you can still see as sensible to continue our discussion; I will just not be always quick-responsive (due to the pressure of other things in my life).
At the moment, I again show a short conversation with ChatGPT, to save some time.
My question to ChatGPT (without any preceding discussion on this matter):
Are there claims on historical events to which we can attach the probability 1 absolutely, or is it reasonable to always leave some, maybe astronomically tiny, room for doubt?
(Abridged) answer of ChatGPT:
In history, it’s generally considered unreasonable to attach a probability of 1 (absolutely certain) to any event, even when the evidence is overwhelming. The reason lies in the nature of historical knowledge and the limitations of evidence, interpretation, and human understanding.
Here’s why leaving room for doubt is usually seen as reasonable:
…..
A Historical Example:
Take a well-established event like the moon landing in 1969. The overwhelming evidence — from photographs, videos, eyewitness accounts, and the physical samples brought back — makes it extremely unlikely that this event didn’t happen. However, even here, some might argue (though it’s a tiny fringe view) that there’s still room for doubt due to the possibility of new, undiscovered information or a misinterpretation of existing data.
In summary, while it may be reasonable to be highly confident about some events (such as the moon landing or World War II), the idea of leaving even a tiny room for doubt is generally a good practice in historical analysis. History tends to be about probabilities, not certainties, because new evidence or perspectives could always emerge.
Neil, at this moment I would just ask if I understand you correctly that you find such an answer unreasonable (for some reason that you think you have clearly explained) …. ?
Hi Petr and thanks for replying. I thought I had lost you for a moment there.
I find the answer inadequate. It is “reasonable” in that it is making a “reasoning” case, but its assumptions and starting position are not those of the real world.
Carrier speaks of there “always” being room for some doubt but that is only so in the theoretical world of mind-games and is not part of the real world. Here is where we have the difference between verifiable facts or events and hypotheses about them.
Take a court case, for example. The jury has to prove beyond “reasonable doubt” the guilt of a person for a murder. There can always in the decision be room for the infinitesimal slither of doubt that for practical purposes leaves no “reasonable room for doubt”. But the murder itself may be a verifiable fact of absolute certainty, all other things being equal. That the accused’s fingerprints are on the knife is a verifiable fact and there is not even the slightest room for any doubt (not in the real world). Whether the accused further stabbed the victim is initially open to question — hence the court case.
In the real world there is not the slightest room for doubt about events that are clearly verifiable. — If I am caught on camera running a red light and get a notice for a traffic violation complete with photo of my car and the red light, there is no room for even the tiniest doubt about what happened. In theory I can claim that someone with a hatred against me in the Traffic Bureau hates me and doctored the photo but I am only playing unrealistic mind-games in desperation if I think so.
Carrier raises possible doubts, extremely unlikely doubts, alongside the world of reality — he would say that we have to admit the possibility of some unrealistic explanation. His examples that he gives in Proving History of such doubts are in fact so unrealistic as to be nonsensical in my view. (I can provide/quote examples if you wish.)
In the real world is there any room for any doubt, however small one might like to measure it, that the Archduke of Austria-Hungary was assassinated in 1914 and that World War 1 followed? There IS room for some doubt, small though it might be, that the assassination was in some way responsible for WW1, but there is no room in the real world for the fact that the assassination happened and that WW1 happened. (Historians will debate why WW1 happened, and even the nature and extent of WW1 in some ways, but the event itself is clearly verifiable. No room for any doubt unless one lives in an unreal theoretical world that has no relevance to reality.)
Is there any room for doubt that Hitler became Chancellor of Germany in 1933 and WW2 followed? Same as above. But to justify his use of Bayes Carrier has to say that even those events have SOME room for doubt, however small. I disagree and say that historians do not use Bayes to verify such events. There is NO room for doubt about many/most of the events used by historian in their debates about hypotheses and interpretations.
In my view Carrier has overreached. It is the same with his On the Historicity of Jesus. He presented too many new and untested arguments and thus opened himself to unnecessary criticism. He would have made a stronger case by focusing just on a few core arguments, in my view.
It is the same with Bayes — Carrier has attempted to apply Bayes to too much. He even lists about a score of historians whom he says use methods that can be reducible to Bayesian methods. I suspect Carrier never actually read many of those historians with any care because I can provide quotations from several of them that flatly contradict Carrier’s claim about their methods. If Carrier had limited himself to using Bayes to test hypotheses, that would have been enough and would have made his point. But he has gone beyond that attempt and asserted that “all claims” by historians — and he means to include what others deem to be “verifiable events” — have some room for doubt. If an event is verified in the real world there is no room for doubt in the real world. And verification appeals not to probable notions but to other verifiable evidence.
A second point follows from the above.
Just because a hypothesis argues that the existence of Jesus is unlikely (Carrier says at best it is 30%), it does not follow at all that Jesus therefore “probably didn’t exist”. Events like the historical existence of some persons are not subject to probability analysis but to verification. If he existed then it would only mean that more fanciful stories were told about him than about many others. That has no bearing on the question of whether he existed or not. I can remove all the evidence we have for Julius Caesar and look at the evidence of ancient Rome prior to Caesar and conclude that the Roman government was set up in such a way to prevent the rise of a dictator … therefore it is unlikely that Caesar would have become a dictator. But all of the practices of Rome, all their safeguards, all their reliance on a great army to protect them from the overthrow of their ways, made no difference to the fact that Caesar became dictator. We might argue that an unlikely event (probability wise) happened. That’s how history works so often.
Let me try to be more direct than my reply above, and then I want to add one other point.
No. Photos etc are only a small part of the evidence. There is nothing “extremely unlikely about it not happening”. That is simply bizarre as a mere suggestion and it is not addressing reality. There are no doubts at all — not “extremely unlikely”, rather “it did in fact happen”. To say it is “extremely unlikely” not to have happened is to yield even too much ground to the conspiracy nutters. It did happen. There are no doubts AT ALL. (Except among circular reasoning confirmation biased conspiracy theorists — some of whom also presumably believe aliens built the pyramids and Atlantis was once a real place.)
No historian says “it is extremely unlikely that it didn’t happen” or they’d have their audience of fellow historians rolling their eyes in dismay. It sounds like it is being conceded that there may be some very tiny possibility it didn’t happen. That kind of thinking is for abstract theorists who are out of touch with reality.
That’s nonsense. We don’t know about the reality of the moon landing because of scattered pieces here and there of items of data. It was a major grand program involving many thousands of participants to make it happen and millions to make it available to be witnessed world wide. Let the conspiracy nutters go to some other universe. They are not living in reality.
Is that statement really saying that we should leave some room for doubt — just as a tiny tiny possibility — that WW2 did not happen? I don’t know what sources ChatGPT was using but it is simply nonsensical to “leave a tiny room for doubt” about WW2 or any other “verified” historical event. Verified means it’s true. It happened. Historians don’t doubt the events per se. They doubt and debate various hypotheses and interpretations and reconstructions of those events. (I’ve found ChatGPT even makes mistakes in straightforward translations. It is a machine, not a historian.)
Define history. The events are not the probabilities. It is the way they are put together to form a certain type of narrative, or the way they are explained — that is where “probabilities” etc arise.
Now for my other bit that I wanted to mention:
Carrier claims to have discovered the logical “secret” that lies at the heart of all historical reasoning and claims. He has little modesty about this achievement. At the head of his chapter explaining in some detail the Bayes formula he places this quote from a historian as a way of introduction:
Carrier has not discovered it at all. In another place I provided another quote from the Preface of the same book by that historian debunking the possibility that historians could be guided by Bayesian reasoning as some kind of catch-all for what they do.
If Carrier wanted to demonstrate that all valid historical methods or “claims” are at bottom Bayesian he could have done so but it would have been a longer book. He claimed here in comments that he had demonstrated it but I don’t think he stuck around to read my rejoinder that all he did was assert it and then list a score or so of historians’ names as “evidence”. I know several of the historians he listed and I know for a fact that they simply outright reject anything approaching Bayesian reasoning as the foundation of their “claims”. I hope to post something to show that.
If Carrier had really discovered what he claimed it would be well known and discussed in major history journals. As it stands, I don’t think Proving History was accepted by any major academic publisher and was taken up by a publisher interested in debunking Christianity and other myths.
Carrier could not even bring himself to read in context any of my comments and he sure as hell didn’t bother to read posts he left comments on. He is evidently not prepared to defend his thesis except to smugly dismiss all opposition as “Bayesian without them knowing it.” I really expected him to have some interest in engaging with a discussion. I am not rejecting Bayes outright but believe Carrier has simply gone too far with it. But I have now been left with the sneaking suspicion that he has no confidence in engaging other historians with his “discovery” and is a fraud getting high on hits from a lay public who want to debunk Jesus.
How the hell Columbia University awarded him a PhD in history when he has never once demonstrated the slightest awareness of how historians really work and write — sorry, I really am dismayed that he could be espousing a kind of approach to history that the field of historians almost entirely debunked in the first half of the last century.
There seems to be a substantial difference in our backgrounds regarding the thinking in mathematics, and science in general. I already touched on this previously: if a scientist S tells me “your calculations are erroneous …”, then when requested to explain where the errors are S must be able to do this, or to retract the claim; otherwise I find no sense in further discussion with S.
In your case I felt as impolite to simply stop discussing with you (though I tried to do this at one moment afterwards).
Neil, I can see many issues in your texts, and I was hoping to clarify them step by step; I find it counterproductive to try to clarify everything in one go.
Let me now try a (hopefully last) attempt to clarify at least one issue, leaving the other issues rather untouched in this post.
You wrote:
“Does anyone really believe there is a nonzero possibility that there was no Second World War? Or First World War? Or American Revolution/War of Independence? Of course not.”
[I suppose you meant “a nonzero probability” instead of “a nonzero possibility”.]
It might surprise you but I really “believe” this if you want to formulate it in this way. I am thus demonstrating that your “Of course not” is invalid.
So now you can declare me irrational and stop discussing with me; or you can try to concentrate on my following explanation.
First I test if we agree on the following:
Suppose some people tell me that they experimented with a die that was confirmed to be fair by thorough testing, and that one day the following event happened: in the standard 1000 consecutive rolls of the die, a six never appeared.
In this case my reaction would be this:
1/ I am convinced that the event did not happen (concluding that either these people lie, or they did not notice some hidden manipulation at work).
2/ I “believe” that the event has a nonzero probability (though less than 0.000….01 where the number of zeros in the sequence is 80).
If you (unlike me) do not believe that the event has a nonzero probability, then we have found that we substantially differ in understanding the meaning of the word “probability”. But I somehow feel that you might see this situation similarly as me.
Now let us move to the history. I use rather the claim C:”Japan attacked Pearl Harbor in December 1941″, which you yourself used as an example and which is more concrete than “Second World War” or so. Here I say:
1/ I am convinced that C is true.
2/ I “believe” that there is a nonzero (though astronomically tiny) probability that C is false.
Using the standard meaning of probability, I can thus say that the probability I am attaching to C is greater than 0.99999, say, but not equal to 1.
It seems that we differ here, since you attach the probability 1 to C. If we both understand probability and “believing in Bayes” in the same way, then you also agree that by attaching 1 to the probability of C you also claim that you are -absolutely- sure that it is absolutely impossible that any new evidence that might appear in the future can decrease this probability 1.(If you do not agree that attaching 1 to the probability of C has this consequence, then we have found that we substantially differ in understanding the meaning of the words that we use.)
Let us now see how I am coming to my 1/ and 2/ for C.
I have no direct access to historical events, so I can only analyze the current relevant evidence that is available to me.
Ok, I remember hearing the claim C somewhere (maybe first at school); is this a reason to be absolutely sure that C is true? I say no. Now I can look at the wikipedia, say: I get close to the probability 1 but not completely. I study the relevant sources, one after another. There is no single one that gives me the probability 1 by itself; if I want to be completely sure, I must also try to be as sceptical about the sources as I find reasonable.
So I get a row e1,e2, …,en (for n=100, e.g.) of the relevant evidence-items (that I consider independent); for simplicity let us assume that each has the odds 99:1 when C is true, i.e.
P(ei|C)=0.99 (and 1:99 when C is false).
I start with the prior probability of C being true 0.9, say, since I heard it at school, and update this by processing the items e1,e2,…,en successively by using Bayes’ theorem. Of course, I find that the posterior probability exceeds 0.99999, so I look at it as at -virtually- 1, while being aware that I have not reached the absolute 1.
I hope that I do not need to stress that I am not really doing any such calculations explicitly; I just find this description to be a solid, say scientific, model of correct reasoning in such matters.
Can you explain me, Neil, what is your process that reaches the absolute 1? Isn’t it just the same process as I describe, in which you simply round 0.99999 to 1 in your head?
I don’t believe most people “know” beyond any “probability” calculus that World War 2 happened by the mechanical and selective steps that you describe: “heard it at school”, “checked wikipedia”. That’s how we might go about topics of which there is reason to doubt but it is not how we know WW2 happened. Ditto for the moon landing. These events involved mass public participation that has left it absurd to think that “it just may be that new evidence in the future will overturn our belief that WW2 happened.”
Ditto for many cases of murder, or a kidnapping, etc. Of the fact of a crime there is no probability arising. It is a known indisputable fact for what are surely obvious reasons.
I understand how probability works — that is not where we disagree. Where we disagree is on the applicability of probability calculations to the necessarily one-time events of history. That is what the Willer and Willer book explains in depth — probabilities have been misapplied to social sciences in an erroneous attempt to make them look more “scientific”. It is not just me Carrier disagrees with but it is broad fields within social sciences as they have evolved through the last century.
Carrier also disagrees with just about every historian I have ever read who discusses methods and reconstructions of the past. His list of historians whom he claims use methods reducible to probability calculus is misleading — in fact it is a false claim about what many of those historians actually wrote.
I can say that in theory WW2 may be found to have been illusion with new evidence one day, but that is a make believe fantasy and not how reality works. That is a test case example that must make or break Carrier’s approach to historical epistemology.
Carrier is a lone wolf in this question among professional historians (and I suspect a majority of social scientists) from what I can tell. He has boasted of a review of Proving History by Tucker without realizing that Tucker was actually not approving of all Carrier wrote at all, nor of his argument. That Carrier has no interest in even attempting to read criticism of his work without assuming the person is being dishonest speaks for itself.
First of all, I do not have a process that leads me to 1 for, say, that WW2 or an attack on Pearly Har bor happened. There is no mathematical process or conclusion at all.
I have grown up with my father and uncles and aunts and mother telling me stories of their experiences in the war. I see around me monuments and two national days set aside for memorials, alongside plaques of names of the dead. I see the whole of society, people of related nations all observing the same kind of thing, and I see films, encyclopedias, and documentaries and interviews all with testimonies of participants from both sides. I see as a part of this many particular battles, including the attack on Pearl Harbor. I know that the society and world I have grown up in experienced WW2 and were all aware of particular moments such as Pearl Harbor.
I do not have to check out memories of school lessons or wikipedia to establish what is evidently a certain historical event and certain major related events.
Then if you are not doing those calculations explicitly, are you doing them implicitly? I am not doing such calculations at all implicitly let alone explicitly to know WW2 and Pearl Harbor happened.
No, it is quite a different process that I have touched upon above. Carrier says that ALL historical claims — and he includes claims that certain events happened — have some theoretical room for doubt, as you are proposing. I disagree.
Knowledge of past events is sometimes based on probabilities but other times it is simply undeniable reality that can accept no room for any doubt however small. Not all historical knowledge is probability-grounded. Most of the events are known to have happened, full stop.
Carrier, as Aviezer Tucker himself notices in his review, confuses events with hypotheses about events. Carrier fallaciously treats the two as one. Hypotheses seeking to explain the whys and wherefores are susceptible to probability analysis. But Carrier confuses the two things — events themselves and hypotheses. It is not a hypothesis that WW2 happened. That cannot be a hypothesis. That is a fact. I can hypothesize about why it happened, the results, etc.
Now, take a crime scene. There are reports of synagogues being sprayed with neo-nazi slogans recently. They are recent historical events. I haven’t seen them directly myself first hand but I do know they happened for a host of reasons that relate to the way our society works. Hypotheses about who was responsible etc are open to Bayesian reasoning. But the historical acts themselves are not.
A small addition: The example value n=100 in my above model is unnecessarily large; but all this is just a model of correct reasoning, I hope that you do not understand this so that I would feel a need to perform such a process myself. (I have good reasons to believe that it was performed by others.)
I think that a crucial problem between us is that you seem to think that when I do not attach the probability 1 (absolutely) to a claim, then I must have some reasonable doubts. But this is not the case (as my example with no six in 1000 rolls hopefully illustrates).
I think we have covered this. I do understand what you and Carrier mean by 0.00000000000000[etc]1 doubts. It is a theoretical view of the data and not a “real world” one. But it is a theoretical view that I (and other historians) do not believe applies in reality of empirical data. Ditto for police coming across a crime scene or a suicide or some other tragic event they are called on to deal with. It is generally very clear what happened in the past (I know they are exceptions, of course, but by and large) and the fact of a certain crime is simply not in doubt — theoretical applications are irrelevant and not part of the real world. They do not apply to contingent one time events. Carrier would say the police do Bayes “in their heads” to interpret the evidence. Very often such Bayesian calculus can be done in retrospect but that is in retrospect and not how we always know what happened in any kind of historical one-time event.
In short, I believe anyone can see from my record of using Bayes and defending Bayes for testing hypothesis is sound. (I have pointed to one of several instances here.) I have in the past defended Carrier’s approach overall. Trying to put forward cases asking me to decide on options to prove to me that Bayesian thinking is standard and valid (even for me) misses the point I have been trying to make. I have criticized Carrier for not using Bayes enough in the case of what I consider ill-informed claims of his.
But “Did WW2 happen?” or rather, “How Do I know WW2 happened?” or a police officer coming across a scene of multiple stabbing deaths …. does not need to use Bayes, in fact in none of these situations is Bayes used except perhaps in the first case for someone who has lived under a rock. “Did X happen?” when asked of WW2 or a murder scene are not Bayesian questions — they are not hypotheses to be tested or confirmed in some way. They are known empirical events that direct and indirect evidence confirms in abundance.
We can and do confirm and verify things all the time without Bayes — and without even using Bayes at a subconscious level. There are situations where certain kinds of data can be established as verified without appealing to Bayes.
The review of Carrier’s book whom Carrier has sometimes appealed to actually stated point blank that Bayes cannot be used in the way Carrier uses it, and he further pointed to where Carrier has confused data with hypotheses. But because he explained in fairness and honesty Carrier’s reply to this point, Carrier has all too quickly assumed he was agreeing with him. He was not and in fact let his disagreement stand.
Tucker let Carrier’s reply speak for itself without comment. But on the other hand, where Tucker could find a point to praise in Carrier’s book he certainly did so. But Tucker has not changed his mind about the epistemological status of historical knowledge in his publications on history since he read Carrier’s book.
Carrier is trying to reduce history to a documentation or identification of “what happened” on the same level as science (only not as exact as science — being different only in degree of exactness).
As Tucker himself noted in one instance, Carrier confuses verifiable events with hypotheses. Tucker pointed to one glaring example of Carrier doing this in Proving History. Carrier’s whole approach to history is in fact one such confusion.
Were it not so, were Carrier’s “discovery” of the heart of all historical reasoning (as he claims it to be) were valid, he would be the hero of history departments everywhere.
Bayesian analysis tells us that Carrier “very probably” does not have the magic bullet he claims to have. 😉
Even though it can’t be proven that Jesus was historical or not, like Robin Hood or King Arthur (once thought to have existed and today carry no weight whatsoever as being historical), maybe by weighing Jesus with the likes of such figures of the Roman, Egyptian, etc. gods we can get rid of the fallacies that now exist within the religions of the world. I welcome all research into such religious matters.
Re: Point 1
There are too many historical possibilities for the origin of Christianity to fit neatly into any framework suitable for statistical analysis: from the inerrancy of the gospels, through Jesus as composite character, to conscious fraud by some Flavian, with many offshoots along the way. I don’t think it’s even clear what the “minimal historical Jesus” would really consist of – e.g. how many red-letter passages in the NT would have to be authentic? 95%, 51%?
As I understand:
a) The core question for historians is not simply whether Jesus existed but rather how Christianity originated. Investigating the actions of Jesus and the subsequent activities of his followers is crucial for understanding the religion’s development.
b) Even if Bayesian analysis suggests a high probability of Jesus’ non-existence (or existence), the outcome remains inconclusive. The historian gains no substantial insight into historical events or the factors that shaped Christianity.
Argument 1:
• Degraded Focus: True historical interest lies in understanding the origins and evolution of Christianity.
• Limited Scope: A simple “yes” or “no” answer to Jesus’ existence provides minimal information about the complex historical processes involved in the religion’s formation.
Argument 2:
• Lack of Practical Application: Regardless of the probability assigned to Jesus’ existence, the historian remains uncertain about the actual historical reality.
• Unhelpful for Historical Reconstruction: Bayesian probabilities, even if seemingly strong, do not provide concrete evidence or insights to guide historical interpretation.
• False Sense of Certainty: The numerical probabilities can create an illusion of certainty that may not accurately reflect the complexities of historical evidence.
Caveat emptor:
“a) The core question . . . Investigating the actions of Jesus and the subsequent activities of his followers…”
To date, however, investigating the actions of Jesus—like the “Jabberwocky” poem by Lewis Carroll—has been one of literature’s epic bits of nonsense 🙂
I don’t think the scholars who do this are engaging in “nonsense”. I think they are genuinely trapped in thinking that they are being objective. The right approach is to try to shed light on the areas where they blur the lines and shapes of what they are investigating. To this end, I see the work of Stephen Young as particularly useful. I posted about his work some years ago but since then he has produced more along the same lines: https://www.academia.edu/73040566/_Lets_Take_the_Text_Seriously_The_Protectionist_Doxa_of_Mainstream_New_Testament_Studies
I am writing about historical methods and historical knowledge because I think not many biblical scholars have thought through these kinds of questions. Even some “nonbiblical” historians tend to sidestep examining face to face their methods. I began my little journey of studying the nature historical knowledge and methods used to build historical knowledge over 20 years ago, initially prompted by the questions raised by Earl Doherty. How do we know anything about the past? Bart Ehrman attempted to answer that question in his published tirade against mythicism but he opined shallow notions from the top of his head and clearly had never taken the trouble to study the question seriously. It’s a serious question and most of what I write, if not everything related to this question, is informed by reading works on the philosophy and methods of historians, past and present.
If you’re already familiar with the following, then just ignore it; but it seems to me that a lot of space is devoted to arguing the difference between what, in the field of legal practice, is considered “reasonable” vs. “possible” doubt. Here is the/a standard jury instruction on the issue:
“Reasonable doubt is present when, after you have carefully considered and compared all the evidence, you cannot say you are firmly convinced of the truth of the charge. Reasonable doubt is a doubt based on reason and common sense. Reasonable doubt is not mere possible doubt, because everything relating to human affairs or depending on moral evidence is open to some possible or imaginary doubt. Proof beyond a reasonable doubt is proof of such character that an ordinary person would be willing to rely and act upon it in the most important of his own affairs.”
In literary studies, they would probably dismiss such an instruction as being “normative” (a pejorative word in the literary world), but that’s the difference between theoretical and practical concerns. History, it seems to me, must at times distinguish between the two in reaching conclusions.
Carrier goes beyond “reasonable doubt”. He will accept that it is “reasonable” to accept as true information that has a very slim chance of proving to be false. The converse, also, of course. Petr here has made the same point.
One might say that a court is testing competing hypotheses of guilt or innocence. It is not, generally speaking and barring the obvious exceptions, testing whether a crime or injury has been committed. Carrier conflates the two: hypothesis and the event about which the hypothesis argues a point, treating them as having the same epistemological status.
I plan to respond later (in a few days) in our thread above, with an attempt to come to some common conclusion on at least one point. Here, I just want to ask about the meaning of your
“Carrier goes beyond “reasonable doubt”. He will accept that it is “reasonable” to accept as true information that has a very slim chance of proving to be false. The converse, also, of course . Petr here has made the same point.”
I apologize but I don’t understand exactly what is meant by it. You might use the following example for illustration.
Let’s say that the evidence of a crime includes fingerprints that match those of person A. If I just make a remark that there is an infinitesimally small non-zero chance that these are someone else’s fingerprints, am I saying, according to your understanding, that it is reasonable to actually consider this possibility in the investigation?
I really simply do not understand how you have meant what I quote above.
(I am just afraid that you might have got a strange understanding of what I am really saying; this was already indicated by your ironic “Pearl Harbor was probably attacked by Japan” and other such remarks that have led me to an impression that you totally misunderstand the notion of infinitesimally small probability …)
So, please, explain the meaning of the quote; I might respond to it in my planned later reply in that other thread. Thank you.
I mean that the probability of the Pearl Harbor attack not happening is so small that it is reasonable to believe that it did happen. In Carrier’s words, Bayesian calculation that it did not happen is so small that it is reasonable to overlook the smallest possibility (let’s say zero point a million zeros after it before the one) that it did not happen. I understand you have been making the same point when you have asked if there is any unreasonable conclusion that follows from Carrier’s approach.
In short, I mean what the judge means when he tells the jury to come to a decision “beyond reasonable doubt” about, let’s say, “the hypothesis of” guilt.
It is entirely reasonable to use Bayes to weigh up hypotheses.
We saw recently how Carrier is able to use the online medium here to avoid having his view of the status of historical knowledge tested. The following passage from the historian Richard Evans in Lying About Hitler is appropriate:
I had thought you were using the WWII example to say that we don’t need to quantify our doubt about some facts of history, because they are beyond question. The phrase “as a practical matter” came to my mind. When that phrase is used, it means we will not accept argument over this issue; it will be taken as a given. This put me in mind of the legal standard of BRD. The law says to the jury, when you reach this point, stop with theory and accept this as the practical conclusion. One can always make a sophistical argument, but as a practical matter, would anyone ever be so foolish as to require a quantified (Bayesian) analysis of the likelihood that WWII occurred? So it had seemed to me that you were arguing for a similar practical rule applicable to historical conclusions and assumptions. Here you seem to say that hypothesis and event should not have the same epistemological status, and that Carrier conflates the two, and I’ve lost sight of the example, but are you not saying that certainty is more readily apparent and may be assumed in some instances than in others? If so, all I was saying is that one basis for making a distinction might be the nature of the proposition, one being more theoretical in nature than another more practical one, if that makes sense. We would need to move to examples.
I suppose Neil will react, but I just wanted to recall that the debate is on Carrier’s axiom 4 recalled by Neil in https://vridar.org/2024/11/24/jesus-mythicism-and-historical-knowledge-part-3-prediction-and-history/
Neil takes this axiom as invalid, though he has so far presented no single instance where this axiom has led Carrier to an unreasonable conclusion or so.
I am afraid that it is Neil who draws unreasonable conclusions from this axiom (in my opinion he is not really contemplating what an infinitesimally small non-zero probability actually is) and then ridicules them …
I am sorry if I have seemed to have made unreasonable conclusions. I was not aware that anything I wrote could be construed that way. What I have tried to make clear is that Carrier is referring to both events and hypotheses in his term “claim”. They are not of the same type of knowledge.
I am sorry you feel that way, Petr, about my responses. I regret not coming across more cordially the way some others seem to find they can so naturally.
Agreed, Carrier does not draw “unreasonble conclusions”. He argues reasonably. We can disagree with another’s reasons, however, can’t we? What does concern me is that my own conclusions appear to be “unreasonable”. I was not aware I was ridiculing anything, by the way, but I stand to be corrected if reminded where I have done so.
As I tried to point out in a response to Clarke, I believe there is some kind of knowledge and belief that we take on board without any reference, even subconsciously or subliminally, to what are in effect mathematical calculations.
I disagree that we can say that there really IS a chance of, say, 0.0000… and add a trillion trillion zeroes after that decimal point before we reach the one, it is really unacceptable to say that there is that slimmest of ridiculously small chance that we are wrong about there having been a WW2 in history.
It is not because I can’t imagine how small that fraction is. It is because I don’t believe that applying maths at all, any probability measure at all, is valid to justify our belief in certain kinds of knowledge. Carrier himself acknowledges this much by mentioning that knowing you can see your hand in front of you is beyond probability confidence measures. I simply extend that same point to other kinds of knowledge. That is where Carrier and I disagree. (Carrier was simply flat wrong to accuse me of saying that I was equating science and history with degrees of certainty — I was trying to say that they are not to be compared as a method of acquiring certain kinds of knowledge. I felt he was the one being unreasonable and less than honest by refusing to read my context and post where I explained all of that.)
I admit I have been influenced in large measure by Willer and Willer who set out why the social sciences cannot be measured through probability assessments or any kind of mathematical measure in the same way the sciences can.
I know I misread some arguments at times. I realize that one argument in particular was not about the probability of an event happening but about the probability that a certain event would explain the evidence we have. I tried to explain that such reasoning is actually circular. In the case of mythicism, Carrier posits a certain myth theory that comes mostly from Earl Doherty and that was formed as a direct result of the reading of the sources. So I simply cannot see how it is valid to claim that the myth theory can “expect” the kinds of sources we see.
In the end, following Willer and Willer’s “Critique of a PseudoScience” as it used to be common in the social sciences, I came to the realization that arguing for the probability of a historical event or person is simply irrelevant to the question of whether such a person or event really existed or happened. At best, all Carrier’s method can tell us is that if Jesus existed then he was different from other figures in multiple ways — a conclusion that should also please apologists.
There is a far simpler conclusion to draw from the evidence about Jesus. The Jesus in our sources is a literary-theological construct. They tell us nothing about a historical figure.
I fear nothing I have said here or earlier will satisfy you. I have been mulling over in my head trying to grasp what it is I do not understand and what points you are trying to help me understand. I wish I could do better.
OK. I thought Neil was saying some claims have a zero probability of being false, and I was simply identifying this position as a practical one, similar to the ones used by courts of law. But thanks for your input.
I am saying that some types of knowledge and conclusions do not involve mathematical weighing of odds. By insisting that everything, all knowledge, is reducible to mathematical odds is not valid, in my view. Much of it is, But there are some things we know differently.
I can only respond very briefly at the moment. Your sentences are very difficult for me to parse, Neil. You seem to be arguing against some mysterious person M who says that we need to do some mathematical calculations before we can take the Japanese attack on Pearl Harbor in December 1941 as a fact, expressed by a proposition (a claim) to which we attach confidence 1 in practice.
Who is this M who says such a foolish thing? And where does M say this? Or am I mistaken in what you are trying to express?
I have never said that. I have always tried to make it clear that I acknowledge some kind of quantitative process is at a subliminal or subconscious level — as Richard Carrier says: lift the hood to examine what is your hidden reasoning process; see what is happening beneath the surface.
(Richard Carrier simply asserted that my reasoning in a recent was Bayesian without sticking around to defend his assertion. A sensible retreat on his part, in my view. I disagree with what Carrier says is “happening beneath the hood” without our realizing it.)
So I am completely confused. Why do you say “I am saying that some types of knowledge and conclusions do not involve mathematical weighing of odds.” as an answer to Clarke? I thought that you were trying to explain that you argue with somebody who says that all types of knowledge involve mathematical weighing of odds …
I have since posted Carrier’s explanation of how maths enters our thinking — https://vridar.org/2024/12/13/the-folly-of-bayesian-probability-in-doing-history/#comment-330763
I agree with what Carrier says there but not with his blanket application to what he means by “all historical claims”.
To try to be clearer — and I apologize for not being clear nearly often enough — I am considering doing another post in which I will attempt to present both sides of this question as simply and clearly as possible. I really need to be sure I am understanding your position and I may use some of what you have written in these comments if that is okay with you.
I am sorry that I can only be extremely concise here these days. As someone with a mathematical background (I am still doing some research in theoretical computer science), I very much lack clear definitions of the crucial notions and axioms (hidden assumptions) that are used in the discussion here. I am aware that a standard reaction to this is that history is not mathematics, but I often find this as an inappropriate excuse.
I have tried to first concentrate on Carrier’s axiom 4, leaving all other issues aside. I can say that I find this axiom very reasonable. In my personal model of thinking, say, we attach a certain (level of) confidence to each claim; in our context, by a claim I mean a proposition on our history that can be actually true or false. The claim C:”Japan attacked Pearl Harbor in December 1941″ is just one very specific
example.
This confidence is often expressed in terms of probability (taken from mathematics), and it then should obey the standard axioms accepted for probability; therefore the confidence attached to a claim can be naturally expressed by giving an -interval- inside [0,1]; a precise quantification is out of question here.
One consequence is that the probability of each “scenario” that we cannot prove to be impossible has probability (i.e. confidence) bigger than zero though it can be completely negligible (like, e.g., the probability that we have got 100 times six in 100 consecutive rolls of a fair die).
In this situation I can only say that my confidence in the claim C is indistinguishable from 1, while I feel not justified to say that it is 1 absolutely.
But in practice, I surely feel justified to handle the claims in which I have confidence indistinguishable from 1 in the same way as other people who attach the confidence 1 absolutely (in their model of thinking that I do not actually understand, and which looks awkward to me from the mathematical point of view).
Unfortunately, I have to finish this post now.
I’ll play the devil’s advocate for a moment and agree with everything you wrote in the above comment and accept all that you imply or understand here about Carrier’s position.
Let’s enter the realm of history wars where historical debates abound.
Historians on each of the warring sides will pull out from their research sure and certain “facts” that happened and that support their particular argument. Those historians do their homework with all diligence. As a result, all of those historians, on all sides, produce “facts” (I really prefer the term “events” but I’ll leave that aside for now also) on which they each declare with a confidence of 1 that they are true facts that really happened. All sides agree that the facts happened — they all agree with a confidence of 1 that they are true, even though to be mathematical purists they might in the recesses of their minds acknowledge that there is a slimmest of margin of possible theoretical doubt, so small that isn’t enough room on a page to write all the zeros.
So we have a history war raging but all the historians agree on the “facts” that they are arguing about. They all agree – on both sides of the debate – that the same events happened.
So how can they be fighting? Where is the war? It is, of course, in the meaning they see in those “facts”. It is in their respective understandings of the significance or consequences or causes of those “facts”. The battle lines are drawn up over how each side “explains” those facts.
Those explanations, Carrier explicitly says, are in fact “theories” that can be tested with Bayes’ Theorem (p. 48 of Proving History) As testable theories they can be used to predict what one will find in the sources or even what is the likelihood of a certain event having taken place in the past.
Willer and Willer demonstrated that such a claim is misplaced empiricism. Explanations in history are not of the same kind or order as mathematically grounded theories in science. Explanations of the evidence or events/facts in history cannot be predict what one will find in the sources or what a historian will discover happened in the past.
What happens in historical thinking is actually the reverse of that kind of prediction. If a particular piece of evidence if found then it will either strengthen or weaken slightly the Bayesian estimation that such and such evidence would be found — but BT cannot predict that find.
The explanation of the sources, and explanations of events/fact, can never predict other sources or “facts of history”. It is new discoveries in the sources or historical events that will oblige the Bayesian to modify their calculation slightly.
This is why I have tried to argue that a Bayesian description of the sources cannot predict the likelihood of Jesus having existed. If we say that our sources allow at best a 30% chance that there was a historical Jesus, all we are really saying is that if a historical Jesus existed he was 70% different from other figures in history. That would not be news to apologists. Rather, is there any major figure of history who is ever like another or predictable — without resorting to circular reasoning?
I wanted to sort out Axiom 4 first, as I stated explicitly. Your reaction starting with
“I’ll play the devil’s advocate for a moment and agree with everything you wrote in the above comment and accept all that you imply or understand here about Carrier’s position.”
again extended the suggested topic and provided again a text that has nothing to do with Axiom 4. Your response simply repeats in different words what you have been already saying so many times, as if you think that I am very slow in grasping all this 🙁
I was hoping that you would explicitly admit that you cannot provide any objective reasons that would allow us to conclude that Axiom 4 is indeed somehow unreasonable. The substance of your criticism can be illustrated by your expressions like this on the moon landing in 1969
“No historian says “it is extremely unlikely that it didn’t happen” or they’d have their audience of fellow historians rolling their eyes in dismay. It sounds like it is being conceded that there may be some very tiny possibility it didn’t happen. That kind of thinking is for abstract theorists who are out of touch with reality.”
Of course, it would be ridiculous if a historian says this in a normal context of a lecture or a publication or so. This is what I meant when saying that you are drawing unreasonable (I mean unjustified) conclusions from Axiom 4 and “ridicule” them. (It was not an emotional complaint of mine; I understand that this could be misunderstood because of my bad English.)
If you think about it deeply, you are just saying how we (here historians) use our language in various contexts. I agree that the expression “extremely unlikely” usually means that the probability (or confidence) is still distinguishable from zero; but this has no relevance to Axiom 4.
I myself find your “axiom” saying that there are (non-tautological) claims on what happened in the past with which we can be -absolutely- certain a bit awkward (also because you have not provided a method how to clearly separate those claims from the other claims).
But I admit that I have no objective means to invalidate your axiom. I find any further discussion on this (Axiom 4 against your axiom) a bit esoteric, having no practical relevance. Do you agree at least with this? I mean sincerely, not by playing devil’s advocate. If yes, then we can proceed if you wish (to more substantial issues in your texts).
I wrote two lengthy replies but then I remembered that you indicated at one time I write too much 😉 … So I have removed them. Instead, I’ll try to leave room for you to do most of the talking:
What in your view does Axiom 4 say and where did my reply miss the point?
What, exactly, was the point in my post, in your view, that was a mere repeat of what I have said so often before?
What do you understand my specific explanation as to the fundamental problem I have with Axiom 4 to be? I believe I have stated it explicitly several times. I do not understand, therefore, your question now. You must know what my objection to Axiom 4 is?
I don’t think you want me to repeat it? 😉 But it might help me if I could be assured you understood my objection to Axiom 4.
Why is that example unreasonable? I thought the whole point of Carrier’s claim is that there must ALWAYS be some room for doubt. I have quoted Carrier saying that, iirc. Yes, I can understand that saying it in a public lecture would sound bizarre — but is that not exactly what Carrier is saying must necessarily happen, at least “beneath the hood”?
Now I am confused by the language. Is the probability that there was a WW2 exactly 1 or is it just a smidgen less than 1? I mean theoretically. If it is 1, even theoretically, then there is no probability ratio involved at all.
This is not a game with semantics. But I would be repeating myself if I explained why in a lengthy reply.
Well I did explain how I know WW2 happened. So I think your objection is misplaced. If you want me to give you a mathematical answer to how I separate true from false claims — at least in all cases — I cannot. If that means in your view I lose the argument, so be it.
You will have to explain to me what you understand to be “my axiom”. I’m lost here, sorry.
Now I am totally lost. Can you be specific. What exactly did I write that you see see has no practical value?
We seem to be from different planets 🙂
Let me bring a (vaguely related) quote by J.L.Savage from wikipedia
“… as to what probability is and how it is connected with statistics, there has seldom been such complete disagreement and breakdown of communication since the Tower of Babel. Doubtless, much of the disagreement is merely terminological and would disappear under sufficiently sharp analysis.”
I am not sure if you are interested in a “sharp analysis” that first clarifies our terminology, so that we can find a common language in which we could really discuss afterwards.
I suppose you agree that we talk primarily about “epistemic probability”, measuring the subjective degree of belief of the individual assessing the uncertainty of a particular situation, in our case of a (simple or complex) claim on what happened in the past.
You yourself surely use your subjective scale of such epistemic probability (I could recall your claims evidencing it, but I hope it is not necessary). In your model you attach the -absolute- 1 to the claim that Japan attacked Pearl Harbour in December 1941.
In my model, my confidence in this claim is derived from the current evidence I have at my disposal. No piece of that evidence gives me the absolute 1; the combination of such pieces cannot give me the absolute 1 in my model, it “just” gives me a value that I cannot distinguish from 1. This does not mean that I would be thus saying “Japan probably attacked Pearl Harbour in December 1941”, “I have a smidgen of doubt that Japan attacked Pearl Harbour in December 1941”, or so.
You seem to be not able to accept that other people find such a model with no absolutes (like, e.g., also Steve Ruis responding in Part 1) much more elegant for good reasons, while you are unable to give a single instance where this model can fail in practice.
(Maybe a mathematical example might help. If somebody gives me the record of 1000 consecutive rolls of a fair die, I have no smidgen of doubt that I find a six there, but I still do not attach the absolute 1 to my confidence.)
So can you just accept that your critique of Axiom 4 is purely subjective?
On the other hand, I am aware that I have no means to show you that your model with absolute certainties is somehow incorrect (similarly as I could not show this to somebody who would claim the absolute certainty that a six occurs in 1000 consecutive rolls).
So can we at least agree that this difference in our mental models is nothing crucial in practice, and it thus should not prevent us to continue clarifying the terminology. Yes, or no?
No.
I would like to clarify terminology, but see my next point…
We are certainly talking past each other. I realized that most fully when in your earlier comment you said something I wrote was a bit esoteric with no practical relevance. On the contrary, everything I have said has the utmost relevance to how historians understand what they are doing and how we understand history.
I think we must have quite different notions of what “history” even is. There was nothing impractical or esoteric about anything I wrote.
I asked some specific questions in my previous comment and I wish you had answered them directly — with a Yes or No as you ask of me.
This reads to me like a word game. You don’t want to use the word “probably” because that sounds bizarre, as if the probability is significant. But in technical fact I think you do mean “probably” in the sense that we have been discussing, and in the sense Carrier means it. Even if the probability has a trillion zeros before it’s final 1 it is still probability.
You ask me if I can point to where it makes a difference in the real world of what historians think etc, even though I have answered that same question several times now. I don’t understand why you do not accept the answer I have given as the explanation of what I am arguing.
Historical events are not statistical events, nor is our knowledge of many of them probabalistic (except in hindsight, perhaps). Not even with a trillion zeros to the right of the decimal point before the 1. There is a place for probabilistic reasoning in historical reconstruction but historians ditched the empirical view of history that Carrier is espousing many decades ago. Not all knowledge of the past is probabilistic. You are entitled to disagree.
I feel like my words are being twisted. I feel every reply of mine is only half read and half accepted. I suppose you feel the same about my replies.
Our understandings of the nature of historical inquiry are very different. I cannot accept Carrier’s view of history and my reading leads me to think most historians today could not accept it either.
There is nothing probabilistic at all, even in theory, about many kinds of historical events. Carrier certainly thinks it matters that “every claim” (his words) is probabilistic in order to argue against the historicity of Jesus. So even in Carrier’s view the question does have real-world implications.
Historical events are not statistical, and our knowledge of many of them is not a matter of probability. Statistics can be derived from certain historical events, but that is an opposite process from the one Carrier is meaning — and unfortunately Carrier does confuse hypotheses and verified events themselves. I mentioned the Willer & Willer book. The same points have been made in other publications about the misapplication of mathematics and probability to contingent events of the past. Maybe some mathematicians look askance at the humanities and cannot agree.
Thank you for your clear answer “No”. Hence my confidence that any further discussion here makes no sense has reached a value that I cannot distinguish from 1. Best wishes, Neil.
I’ll try to be as clear as I can be without making a bigger muddle of things. I understand that Carrier argues that in coming to certain decisions or beliefs our minds can tend to weigh up probabilities for and against this and that — NOT explicitly mathematically, but at a subconscious gut-feel level of different “weights” for this or that piece of evidence.
Carrier’s own words are these:
I agree with all of that.
What Carrier is arguing — and what I have also defended and I still do defend in many cases — is that the advantage of Bayes, the advantage of stopping and making the effort to assign some ball-park numbers to our “gut feelings” of weights of evidence, etc, is that it helps us process our thinking more clearly, bringing to the surface what we only half sense or may even be in danger of overlooking. I think that is a helpful thing for many processes.
But again — I believe Carrier is stretching things too far when he includes even confirmed and contingent events of the past in the same bucket as hypotheses and opinions and interpretations. He was also speaking less than accurately when he said my recent post belonged to that type of reasoning.
The discussion between Petr and me has come to an end but for the sake of the record I am posting here two comments I had on hold pending what I hoped would be clearer explanations and direct answers to my questions ….. It is evident Petr and I unfortunately do not even agree on “what history is” so for what it’s worth, here is what I had planned for a future time….
Well actually I was addressing Axiom 4 — in “different words”, at least you recognize. I have said repeatedly — to use the same words — that I object to Axiom 4 because of Carrier’s meaning in his word “claims”. He embraces two entirely different types of historical reasoning or thought processes. I could repeat them but I would be using words I have used before:
Under “claims” he includes both verified events AND hypotheses about those events.
By eliding the two types of knowledge, he is attempting to reduce history to an empirical science (different only in degree of accuracy, not in fundamentals of method). His idea of history was rejected by most historians back in the 1950s. Pretty much the entire Social Sciences field, as far as know but I may be wrong, has for most part ditched Carrier’s view of how to reason about human events.
Again, to use words I have used before, I have never said Carrier is “unreasonable” — but I do deny that his view of history is valid. History is not reasonable in the same way science is reasonable. Carrier is very reasonable. But I disagree with his reasons and arguments — and so do most historians from what I read about their methods.
I don’t recall “ridiculing” anything in this discussion, by the way.
I am not sure what you want me to agree to. Do you want me to admit that we can never be absolutely sure about certain events having happened in the past or not? What did I say in my earlier reply that I got wrong as the “devil’s advocate”? I think you are saying I didn’t get your point correct.
I don’t know what you understand by “my axiom”. What is “my axiom” in your view?
What, exactly, in my earlier comment is in your view mistaken?
And the second …..
If it is reasonable to say that there is some small chance (very small chance but nonetheless a real one that is on the most sensitive scale still measurable) that there was no World War 2 then I agree that Axiom 4 is reasonable and if reasonable, valid. If it is reasonable to say that there is some small chance that Joe Biden was not the previous president of the United States then I agree that Axiom 4 is reasonable, and therefore valid.
My “devil’s advocate” comment was meant to say, okay, for the sake of argument I will agree with Axiom 4 with respect to one kind of “historical knowledge”. But that is not the only kind of historical reasoning and it is not all that Carrier is arguing about history or Bayes. Is it really possible to reduce explanations of human events to Bayesian argument and come up with an objective answer that any unbiased person should agree with? That was the principle that justified the fanciful totalitarian societies of 1984 and Brave New World.
My point is that not all knowledge is determined by Bayesian calculation — not even subconsciously “beneath the hood”. Bayes was never meant to be a test for all claims.
Does anyone really believe that all questions can in theory at least be answered with a single correct answer by plugging in the data and applying a Bayesian test to it all — that is what Carrier is arguing in principle. He is saying history is scientific in its thought processes and by realizing the Bayesian nature of that process it is theoretically possible to plug in all relevant information and produce a final (scientific) answer to historian’s questions.
Carrier writes in PH, page 7 (where he speaks of “progress of historical knowledge” in a context that makes it sound as if he means historical knowledge can advance in a way comparable with scientific knowledge):
and on page 48
I find it difficult to imagine that any historian would agree with Carrier. No doubt there are a few, but all that I have read by historians discussing their methods and nature of historical knowledge — and I have read scores of them, including most of the names of any prominence — tells me that Carrier is struggling with an empirical view of history that was ditched soundly before the second half of the last century.
Carrier in the above quote (p 7) does not even understand the nature of bias and objectivity in historical reconstructions and narratives or debates or explanation. His claims are ill informed.