Re: The Conjunction Fallacy Fallacy [WAS Re: Anti-singularity spam.]

From: Ben Goertzel (ben@goertzel.org)
Date: Wed May 03 2006 - 14:59:49 MDT


Damien,

More recent research on categorization has shown pretty clearly that
the exemplar theory well-describes some but not all examples of human
categorization...

Anyway, the exemplar approach to categorization may be a useful
heuristic in many cases, but it definitely does lead to bad errors in
many cases.

The hypothesis that the human mind uses exemplars to define categories
is not really a refutation of the point that humans make a lot of
silly errors regarding uncertainty quantification -- it's more like an
explanation of one reason *why* these errors occur...

Humans are better at forming meaningful and useful categories than
current probabilistic inference systems, but worse at estimating
probabilities regarding relationships between categories based on
data. Presumably future AI's will vastly excel humans in both of
these regards...

-- Ben G

On 5/3/06, Damien Broderick <thespike@satx.rr.com> wrote:
> At 12:21 PM 5/3/2006 -0700, Eliezer wrote:
>
> >Richard Loosemore wrote:
> >>Human minds are designed for immensely sophisticated forms of
> >>cognitive processing, and one of these is the ability to interpret
> >>questions that do not contain enough information to be fully
> >>defined (pragmatics).
>
>
> >Old, old, old alternative hypothesis disconfirmed a dozen ways from
> >Tuesday... Go forth and read the literature before you make up your
> >own interpretations.
>
> Odd how often we see an imputation of the deficiency of an opponent's
> background research. This gets very old, old and tired. I would argue
> that Richard's case is more persuasive than Eliezer's, basing my
> assessment on the old, old but evergreen hypothesis of Eleanor Rosch,
> together with its empirical support, that the human mind functions by
> observing/creating prototypical exemplars of categories. See for
> example, from old, old, old 2001:
>
> http://amor.rz.hu-berlin.de/~h2816i3x/LexSemantik3b.pdf
>
> If Tversky and Kahneman take significant issue with the Roschian
> case, I haven't seen it--but then my damned library is thousands of
> klicks away.
>
> Damien Broderick
>
>



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:56 MDT