From: Eliezer S. Yudkowsky (sentience@pobox.com)
Date: Sat Jun 21 2003 - 20:41:02 MDT
Harvey Newstrom wrote:
> Eliezer S. Yudkowsky wrote,
>
>> The difficulty here stems from two different senses of the word
>> "disagreement". Ideal Bayesians who "disagree about values" still
>> cannot "disagree about facts". That is, having different values does
>> not allow ideal Bayesians to disagree about facts, including the fact
>> of who assigns what values. Perhaps this means that the term
>> "disagreement" should not be used for differing values, and we should
>> simply say that Bayesians may "assign different values".
>
> This is an extremely important point, I think. Most of the
> "disagreements" on the list are not really disagreements. Different
> people have different data or assign different values. Most of the
> facts themselves are not in dispute. This may be the primary root of
> most if not all semantic misunderstandings. If this were recognized
> more often, perhaps more people would act in a more Bayesian manner?
I disagree.
:)
The above point is what I wanted to warn against when I talked about the
danger of generalizing from ideal Bayesians to humans, because it is
possible to have simplified Bayesians that are missing a kind of
complexity that general Bayesians can have, i.e., nontrivial structure in
computing the utility function. It is possible to "disagree" over values,
not just "assign different" values, but only if you are a certain class of
mind, such that values can depend on probabilistic external facts or
probabilistic approximations of computations.
-- Eliezer S. Yudkowsky http://singinst.org/ Research Fellow, Singularity Institute for Artificial Intelligence
This archive was generated by hypermail 2.1.5 : Sat Jun 21 2003 - 20:52:45 MDT