[notaurus]

> The next time somebody claims that the human brain operates according to bayesian logic, simply ask them what prior probability they (currently, until the next update) would assign to that claim.

A = a person is able to (correctly or not) quantify their priors

B = brain operates according to Bayesian logic

P( A ) ≈ P( A | B )

[trashtester]

So you're saying that P(A) is almost independent of B, as in

P( A ) ≈ P( A | B ) ≈ P( A | ¬B )

?

I suppose my argument is that if someone is using priors that are either a hard 1 or a hard 0, they've removed themselves from the ability to use Bayesian logic at all in any situation where data points in the opposite direction of their priors.

While you can still call it "Bayesian" if you insert a prior of 0, I think such an argument is a direct contradiction of the purpose of using Bayesian logic.

In other words, I would argue that P( ¬A | B ) ≈ 0. Refusual to admit a prior greater than a hard 0 is not compatible with Bayesian logic. Prior probabilities of exactly 0 should be seens as outside of the valid domain within Bayesian logic under most circustances.

[FrustratedMonky]

Maybe we are talking at different levels.

The people theorizing that the brain is Bayesian are not saying that humans do it consciously. Like they can examine priors and making decisions.

It is just that the neurons in the brain update in a way that can be modeled roughly as Bayesian. It happens without us 'deciding to do it', it is just how the brain updates to process the environment. It is happening continually, as we take in senses, and update our internal model.