Is Agreeing to Disagree Honest or Rational?

[mindstalk]

Robin Hanson says no. If you're a truth-seeking rational thinker who understands disagreement theory, and you meet another such, you should come to agree on all matters of fact. The agreement might be on a probability distribution, i.e. agreeing on uncertainty, but that's not the same as agreeing to disagree. If you believe X is true, and another meta-rational believes Y is true, and you meet, something should change.

He mentions Gulliver's Travels, in that the Houyhnhnms agreed too much to seem human; here's a relevant link to the text.

Comments

[wnoise.livejournal.com]

There's actually beeen some work on bayesian reasoning, and how if people start with probability distributions that are far enough apart, collecting more data doesn't necessarily drive them toward convergence -- if you're fairly convinced that X is crooked, while I think he's the bees knees, then it is "rational" for me to consider news sources explaining his crookedness as coming from a vast conspiracy theory, which increases my distrust for those news sources. If other sources are reporting that's hogwash, I'll again do bayesian updating that increases my trust in them. With the right set of data, and starting point, either position can get more and more reinforced.

[mindstalk.livejournal.com]

Huh. Do you have references I could pass on to Robin?
I'm sad to admit I didn't follow the proof in his disagree.pdf and other papers; too much notation I didn't know, it felt like.

I know about that stuff

[(Anonymous)]

That is a literature about if you start with different priors and get the same info, you might not converge in opinion. I'm focused on the situation where you have the same prior and get different info. Robin Hanson

Re: I know about that stuff

[mindstalk.livejournal.com]

Ah. Obvious if you put it that way. I hadn't translated what wnoise said about sufficiently far apart distributions into "different priors".

Re: I know about that stuff

[wnoise.livejournal.com]

Ah, that wasn't clear, and I hadn't had a chance to read your paper yet.

Damien: I don't know of a terribly in-depth discussion, but Jaynes does briefly go over it around page 127 of _Probability Theory: the Logic of Science_