$30k challenge to Interpreter’s “Team Bayes”

[Dr Moore]

https://interpreterfoundation.org/estim ... vidence-3/

TL;DR

Effort to establish statistical independence between a "long book of its size" and dictation of a book of its size.
None.

[Analytics]

https://interpreterfoundation.org/estim ... vidence-3/

TL;DR

Effort to establish statistical independence between a "long book of its size" and dictation of a book of its size.
None.

It's even worse than that. In Episode 1, there were two nearly-impossible outliers regarding book length--the Book of Mormon and Charles Dicnens' Pickwick Papers. This should have been a red flag that there was a problem with the model, but instead he justified his odds regarding page length because "...Dickens was one of those precociously talented, voracious readers mentioned above. He also wrote his first novel in monthly installments as a serial, being paid for each installment, over a span of 20 months. That may sound quick, but it’s got nothing on Joseph’s 65 working days."

After already using 65 days to justify long odds on the book length, he uses it again to lay on another 100,000:1 against it being a fake.

Clearly, the mechanism he used to write a long book is correlated with the one he used to write it quickly. That's obvious, isn't it?

[Dr Moore]

After already using 65 days to justify long odds on the book length, he uses it again to lay on another 100,000:1 against it being a fake.

Clearly, the mechanism he used to write a long book is correlated with the one he used to write it quickly. That's obvious, isn't it?

It should be obvious, if Kyler were aiming for even a semblance of a good-faith analysis. He won't take the feedback, and I can only guess as to why. Maybe he's invested so much time pulling all of his planets and stars together, he's just too proud of his universe to entertain critical feedback.

Also, if I'm not mistaken, the alt hypothesis for Joseph was not 65 days. It would have been 4+ years, for starters, collecting ideas and building narratives in his mind etc. Lastly and it's been said before, but the alt hypothesis must consider the Bible as a starting point for Joseph.

[IHAQ]

After already using 65 days to justify long odds on the book length, he uses it again to lay on another 100,000:1 against it being a fake.

Clearly, the mechanism he used to write a long book is correlated with the one he used to write it quickly. That's obvious, isn't it?

It should be obvious, if Kyler were aiming for even a semblance of a good-faith analysis. He won't take the feedback, and I can only guess as to why. Maybe he's invested so much time pulling all of his planets and stars together, he's just too proud of his universe to entertain critical feedback.

Also, if I'm not mistaken, the alt hypothesis for Joseph was not 65 days. It would have been 4+ years, for starters, collecting ideas and building narratives in his mind etc. Lastly and it's been said before, but the alt hypothesis must consider the Bible as a starting point for Joseph.

Arguably longer than 4 years as he was already renowned for making up stories about the continents indigenous population.

[Analytics]

After already using 65 days to justify long odds on the book length, he uses it again to lay on another 100,000:1 against it being a fake.

Clearly, the mechanism he used to write a long book is correlated with the one he used to write it quickly. That's obvious, isn't it?

It should be obvious, if Kyler were aiming for even a semblance of a good-faith analysis. He won't take the feedback, and I can only guess as to why. Maybe he's invested so much time pulling all of his planets and stars together, he's just too proud of his universe to entertain critical feedback.

Also, if I'm not mistaken, the alt hypothesis for Joseph was not 65 days. It would have been 4+ years, for starters, collecting ideas and building narratives in his mind etc. Lastly and it's been said before, but the alt hypothesis must consider the Bible as a starting point for Joseph.

He set up the series with a prior belief against authenticity of 43-gazillion-to-1. Obviously that wasn't an arbitrary number--it was selected because he was planning on overcoming it in 23 installments. If he were to partly accept valid feedback even occasionally, he could easily be left concluding that the odds against authenticity are only several-million-trillion-to-1. That would be a disaster.

He has no slack to give.

[Gadianton]

Obviously that wasn't an arbitrary number--it was selected because he was planning on overcoming it in 23 installments.

yeah, I'm sure he spent hours running numbers before the analysis to see how it needed to be set up.

His approach as revealed in the introduction surprised me, because I swear I remember him commenting about his upcoming project and mentioning how close of a call it's going to be. I assumed that meant he was going to go the Carrier route -- as you've described it in the past -- where he would fabricate numbers to match carrier's conclusion of something like 60-40 (I can't remember what you said it was). I don't remember Carrier's prior, but I assumed maybe S-R would start at 50-50, and then arriving at 60-40 or something to avoid making the ridiculous spectacle of the Dales.

Is he going to have the Book of Mormon crush the odds for the skeptic, and make it true a hundred billion to one? Or is he going to overcome the huge prior from lightyears afar in outer space, and then come to rest on a knife's edge at around 50-50 or a little better?

I thought what he was going for, was to radically confirm belief for those who already believe, but to make barely more likely than not if you're a hardened critic.

[Dr Moore]

Is he going to have the Book of Mormon crush the odds for the skeptic, and make it true a hundred billion to one? Or is he going to overcome the huge prior from lightyears afar in outer space, and then come to rest on a knife's edge at around 50-50 or a little better?

I thought what he was going for, was to radically confirm belief for those who already believe, but to make barely more likely than not if you're a hardened critic.

Ending at something with one decimal point, like 0.5 or 0.6, would fit the narrative. A solid chance, by no means removing free agency, but enough of a chance that belief is a very reasonable choice, given the evidence*.

* If you count almost none of the "ones" through "fives"

[Gadianton]

god is a mathematician

[Dr Moore]

god is a mathematician

Kyler takes a lot of questions with a smile, but you can see from his responses that he won't change a thing because he knows he's right.

I've seen this before with engineers who work on really complex problems for too long. They fall in love with their solutions, unaware of a fatal flaw in the process.

Here's one line of questions and responses in the comments.

Also, Dr Rasmussen, as I understand it, you've been offered 30 grand if you can:
1. prove, with data, that each probability is statistically independent (if you prefer, uncorrelated or mathematically orthogonal) to each of the other probabilities.
2. Submit these proofs in writing.
3. Pass review by a current BYU statistics or stochastics professor of my choosing.

Were ;you planning to collect on that handsome sum?

Nope! I hope Dr. Moore finds a worthy cause for his excess monies.

(this response is upvoted by Dan Peterson, Kiwi57 and Louis Midgley. Lending their endorsement to Kyler's stiff arm toward incentivized peer review)

To which Daved6 offers the logical follow-up:

Why not? Is it requirement 1, 2, or 3 that would be hard to meet? Or maybe all three?

If, say, the probabilities are not statistically independent then there's no validity to your analysis, correct?

Or

Are you concerned that a legitimate statistics or stochastics professor would not sign off on it?

One would think if your analysis holds merit now, it should already meet the requirements. What do you think fails?

And Kyler responds with a wall of special pleadings, straw-men, and ultimately a long-winded version of "I'm too lazy or too convinced of my rightness, or both" to do the work. (emphasis mine)

You may have missed my answer to noel on this a couple weeks ago, so I'll repost for your benefit:

"As soon as he can explain to me how it would be possible to statistically demonstrate the orthogonality of things like DNA evidence and the presence of Chiasmus in the text, I'll get right on that.

It's a red herring, and I suspect he's very much aware of it. There are cases where independence can and should be demonstrated quantitatively. Such is generally not the case for the historical Bayesian approach adopted by Carrier and others. In the majority of cases, since it would make no sense to try to calculate correlations on the basis of data (what's the process for calculating the correlation between the probability of me contracting COVID and the probability of Dr. Moore having a had a haircut on the last week?), we can instead make informed and careful assumptions about whether two things are independent. This sort of thing is done in statistics quite often, and is generally uncontroversial if done in a responsible and transparent way.

You and your good friends are free to read the FAQ and the essays and decide for yourselves if what I'm doing is reasonable or not. I have little desire to jump through hoops that don't actually exist.

Upon further reflection, the haircut/COVID example is actually an interesting one, because I actually could calculate a correlation for that if I wanted to, because both of those variables could have a workable third variable to structure them around: time. I could try to calculate the probability of both me contracting COVID and the probability of a Moorian haircut for each week in 2021, and then correlate those respective variables.

You generally can't do that when it comes to the probability of observing different classes of historical evidence. What third variable would, say, DNA evidence and Chiasmus vary on that could let me build a meaningful dataset? Is that a question that even makes sense to pose?

Let's pretend for a second that it does make sense. Let's build an imaginary dataset of a couple hundred fictional books that deal with real-world cultures migrating to the other side of the world. For each of those books, I could estimate the likelihood of that culture's DNA being lost to time. I could then estimate the probability that each book would, on the basis of chance, demonstrate that culture's key literary forms. I could then conceivably be persuaded (probably at the point of gun) to calculate the correlation between those two variables.

Hopefully it's clear by now that this is a fool's errand, but let me list a few core issues: 1) I've now multiplied the work required by a factor of at least a couple hundred, 2) it's very unlikely that I could ever cobble together a large enough dataset of comparable works to make that correlation meaningful from a power standpoint, and 3) since those probability values would all be estimates instead of direct observations, the resulting correlation coefficient would be no more valid than the estimates themselves.

Or I could instead conclude that, since there's no logical reason why those things should be connected, it's reasonable to assume that they're independent. If someone ever points out a logical connection that I've missed, I'll be happy to reconsider. But until then it's not something I'm going to lose sleep over."

So let's be clear, for I guess the third time now. It isn't a red herring to demonstrate independence. It's a requirement.

Kyler constructs a straw man by taking one obvious instance of two independent things and making that into "everything" about my $10k challenge. That's not what this is about. It's not a red herring and I've done nothing but act in good faith here.

To be clear: Kyler is NOT comparing the probabilities of independent processes. He's multiplying the outcomes of conditional Bayesian probabilities. Each of his Bayesian conditionals shares one common, correlated, causal factor: the ideas, reading, reasoning and aspirations of one man, Joseph Smith. This fact is evident in the construction of his Bayesian calculations -- in his priors, all of which end up getting multiplied together as he works to overcome 1 in 10^40 odds.

Now Kyler avoids this fact. He presents prior probabilities and acts like they are independent. He gives lip service, but still won't do the work or take the criticism or alter anything no matter what kind of peer review comes his way.

But just because he ignores the fact that his priors all share Joseph Smith's mind as the one correlated, common causal factor, doesn't mean it isn't there. Any trained statistician looking carefully at the construct will see this, flag it, and ask for either (a) data or (b) a carefully reasoned logical explanation why it's safe to neglect. Kyler has done neither.

And so, since each of his Bayesians shares a correlated, causal, common factor (i.e., the ideas, reading, reasoning and aspirations of Joseph Smith) then Kyler is acting in horrible bad faith with his audience and has really hoodwinked the Interpreter crew.

To earn $10k, all Kyler has to do is make the case for independence among his Bayesians through a proper mix of (1) logical argument - perfectly fine in comparing some processes, or more ideally, (2) experimentally gathered data in a controlled fashion, one by one; and then he has to pass those arguments off with a BYU stats professor of my choosing.

I am still baffled about the out-of-hand refusal to entertain this offer. Most authors of academic publications receive no compensation whatsoever, and yet they will do the hard work of taking peer review. Because intellectual honesty and analytic rigor matter to honest academics.

But wait, there's more! To demonstrate even more good faith -- because these clowns seem committed to shooting themselves in the foot and I hate seeing that in slow motion-- I will give Kyler a head start by granting his oft-used straw man about DNA and chiasmus: Kyler, go ahead and take independence between those two processes for free. Now how about all of the rest?

But wait, there's even MORE! Since Kyler continues to dodge the essential challenge here, I will see his "red herring" and do him one better. I will award Kyler half of the challenge reward, $5,000, if he will simply will tackle independence among all of his Bayesian constructs with a reasoned logical analysis explaining why the components of his Bayesian conditionals should all be considered as independent processes from the components of his other Bayesian conditionals. That's it. Now, it will be a challenge to explain why Joseph is not correlated with Joseph amongst all 23 of these conditionals, but if Kyler is convinced, let's hear it. He can skip the controlled data experiments, if that work is too hard or he's too lazy to try, and still win half the prize. And if a BYU professor of my choosing reads it all and signs off on the reasonableness of "independence" in Kyler's work, then the $5k is his.

What excuse will he give next?

[Philo Sofee]

Dr. Moore
Kyler constructs a straw man by taking one obvious instance of two independent things and making that into "everything" about my $10k challenge. That's not what this is about. It's not a red herring and I've done nothing but act in good faith here.

That's your first mistake. Apologists are not acting in good faith to actually work accurately Bayesian probability. Yes, you win the moral war here. Wait a few more posts, and you will win the intellectual one as well. Besides, I could make a terrific video here of this apologist actually DODGING using Bayes accurately on purpose! This could be really, really GOOD...