RE: Metarationality (was: JOIN: Alden Streeter)

From: Ben Goertzel (
Date: Sat Aug 24 2002 - 13:38:34 MDT

James Rogers wrote:
> On 8/23/02 6:57 PM, "Ben Goertzel" <> wrote:
> >
> > I champion conscious reason, but also intuition. I see intuition as
> > unconscious reason, adept at piecing together a huge number of pieces of
> > weak evidence in statistical and creative ways. Intuition can do some
> > things that conscious reason cannot, because of conscious
> reason's limited
> > channel capacity. However, human intuition is also
> inextricably bound up
> > with human emotion, which is also irrational.
> I don't think you are using "intuition" the same way as the pedestrian
> vernacular. There is both rational and irrational intuition, and
> unless you
> specify a definition, you are not being clear.

I'm sure that the common vernacular use of "intuition" mixes up the two.

My sense is very close to the first senses:

-- The act or faculty of knowing or sensing without the use of rational
processes; immediate cognition.
-- A sense of something not evident or deducible; an impression.

Except I would modify the definition to

"The act or faculty of knowing or sensing without the use of *conscious*
rational processes."

because I believe that much (but not all) of unconscious intuitive
processing is actually rational/inferential in nature.

> An irrational intuition/belief is one where the premise(s) have a null
> prior. A rational intuition/belief is one where the premise(s) have a
> non-null prior. While you may not be able to "prove" your
> belief/intuition
> in this framework, you can reasonably assert the rationality IF you can
> demonstrate valid priors for the premises regardless of a lack of evidence
> for the specific position.

There are two problems with this analysis

1) There are really almost NO intuitions in the human mind that have "null
priors" in the sense of not being based on any evidence at all. If there
were, we'd have no way to identify them, to distinguish them from intuitions
based on unconscious evidence.

2) Most of the irrationality in intuition comes in the form of emotional
bias driving thousands to millions of unconscious speculative inference
steps, whose answers are then compiled into a single intuitive impression
presenting itself to the conscious mind

So I don't think that an analysis in terms of prior information really cuts
to the heart of the matter.

> If you've actually evaluated the priors in any type of rigorous
> fashion, it
> becomes "reasoning" and "rational" no matter how weak the association.

What if the priors are numerous but weak so that to really grasp them you'd
need to be able to simultaneously process a million items?

Then your consciousness can't handle the priors, only your unconscious
can.... So the only reasoning your mind can do is the "intuitive" kind.

> In
> essence, "intuition" is short for "talking out your ass" as
> typically used.

Perhaps among the people you associate with, this is the sense of the word

But it is not the dictionary sense and nor it is the sense used by people I
know (which is a broad group: AI theorists, artists, musicians, writers,
deodorant company executives, biologists, and so forth...)

> For all intents and purposes, "intuition" in an AI would be more correctly
> called "low probability reasoning".

It's not low probability reasoning that I'm talking about -- you can have
conscious, deliberate low probability reasoning.

It's reasoning based on the integration of a huge number of weak pieces of
evidence. This is the kind of reasoning that consciousness cannot handle.
In humans. Whether consciousness will be able to handle it in AI's is an
interesting question. I think the answer will be no, for a while. But
post-Singularity I'm not going to venture a guess....

>I think most reasoning in humans from
> null priors is caused by biological biases in the wetware, not as an
> intrinsic function of computing machinery.

I don't agree with your analysis of intuitive reasoning in terms of "null

I think that human intuition is partly directed by emotional and biological
biases, and partly by inferential processes tuned to handle large amounts of
weak evidence in an efficient and pragmatic, parallel-processing sort of

-- Ben G

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:40 MDT