Re: How You Do Not Tell the Truth

From: hal@finney.org
Date: Thu May 03 2001 - 16:49:54 MDT


The part about Robin's paper (http://hanson.gmu.edu/deceive.pdf) that
I have a hardest time understanding is the discussion of common priors.

I'm not sure which of the following arguments is being made, if any:

Argument A:
  If people have common priors, then upon discussion they will come to
  agreement.
  People don't come to agreement.
  Therefore people don't have common priors.
  For people not to have common priors means that they do not seek
  the truth [1].

Argument B:
  If people have common priors, *and use Bayesian reasoning in trying
  to seek the truth*, then upon discussion they will come to agreement.
  People do have common priors [2].
  But people don't come to agreement.
  Therefore they don't use Bayesian reasoning in trying to seek the truth.
  But if they really wanted to seek the truth, they would use Bayesian
  reasoning, or at least try to approximate it, and the conclusion would
  still hold.
  Therefore people don't want to seek the truth.

Argument C:
  If people have common priors, then upon discussion they will come to
  agreement.
  People do have common priors [2].
  Therefore people will come to agreement.
  But people claim not to come to agreement, and believe that they don't
  come to agreement.
  Therefore people are lying about this to themselves and others.

[1] "We are claiming that in the real world, these non-common priors
are best understood as non-truth-seeking."

[2] "And if the process of conception produces an acceptable causal
ancestry relation, then the fact that all humans share a common
evolutionary ancestor makes all humans relevant relatives. So why should
not all humans, including John and Mary, share a common prior?"

It's not clear to me whether the question of whether two humans share
a common prior is an empirical one, or a matter of definitions and
methodology. Is it a matter of fact whether you and I share common
priors? Or is it a question of which assumptions are most useful to
analyze our interactions? (Or something else?)

I also don't understand the claim [1] above relating non-common priors
to lack of truth seeking, unless it is a short-hand for argument B.

Hal



This archive was generated by hypermail 2b30 : Mon May 28 2001 - 10:00:02 MDT