## Richard Carrier and Bayes' Theorem, One More Time

Richard Carrier and I may have found a partial agreement when it comes to the use of Bayes' Theorem (but maybe not).

Take this scenario:

Since there is no objective evidence of such a claim there is no data to work off.

Therefore...

1. Assign a 0% prior and be done with such hypothetical hogwash.

2. Admit Bayes is not the tool to assess such a claim until objective data exists, otherwise we'd have to use Bayes to assess whether Leprechaun's live inside rocks too.

3. Get a team of experts together to agree to a prior probability. Good luck with that!

My friend Grant Shipley responded:

Richard Carrier responded:

To this absurd impossible claim I simply say show me the evidence.

Carrier however, must calculate the prior probabilities first. He cannot say they are abysmal. If he's using Bayes then HE MUST USE BAYES and this means using the math. He must decide between 1 in a billion odds, or 1 in 2 billion odds, or 1 in 3 billion odds, or 1 in 10, 1 in 15, 1 in 20, to 1 in a gazillion odds before he does anything, if in fact he's using Bayes to evaluate this claim. His decision would take some time, since he should be specific by calculating the mathematical odds based on known data (which there are none, and henceforth, this prior probability calculus cannot even be done).

That he doesn't actually use the math means he's not actually using Bayes. He saves a step like all reasonable people do, by dismissing such a claim as impossible and then asking to FIRST see the objective evidence, if there is any.

Take this scenario:

It is impossible that as of this very moment all green highway signs will verbally discuss current events with us, if we stop to talk with them.There are three ways to deal with this in Bayesian terms.

Since there is no objective evidence of such a claim there is no data to work off.

Therefore...

1. Assign a 0% prior and be done with such hypothetical hogwash.

2. Admit Bayes is not the tool to assess such a claim until objective data exists, otherwise we'd have to use Bayes to assess whether Leprechaun's live inside rocks too.

3. Get a team of experts together to agree to a prior probability. Good luck with that!

My friend Grant Shipley responded:

Option 2 is, to me, the correct choice. Bayesian analysis does not expressly include a null hypothesis, but instead is a tool than can be applied, if at all, only once evidence has been accepted that the proposed event/occurrence exists at all.I agreed, which was a step in the conciliatory direction.

Richard Carrier responded:

It's all Bayesian. Your own argument is Bayesian. "Absurdly low prior + no evidence" = "Absurdly low posterior." There isn't any reason to denigrate the logic of your own argument. That's like eating your own foot.Is it?

To this absurd impossible claim I simply say show me the evidence.

Carrier however, must calculate the prior probabilities first. He cannot say they are abysmal. If he's using Bayes then HE MUST USE BAYES and this means using the math. He must decide between 1 in a billion odds, or 1 in 2 billion odds, or 1 in 3 billion odds, or 1 in 10, 1 in 15, 1 in 20, to 1 in a gazillion odds before he does anything, if in fact he's using Bayes to evaluate this claim. His decision would take some time, since he should be specific by calculating the mathematical odds based on known data (which there are none, and henceforth, this prior probability calculus cannot even be done).

That he doesn't actually use the math means he's not actually using Bayes. He saves a step like all reasonable people do, by dismissing such a claim as impossible and then asking to FIRST see the objective evidence, if there is any.

## 0 comments:

Post a Comment