RE: Metarationality (was: JOIN: Alden Streeter)

From: Ben Goertzel (
Date: Mon Aug 26 2002 - 08:30:16 MDT

> Given this, I would then state that rationality is the ability of these
> processes to make use of the BPT in an efficient, objective, and right
> manner, which is to say rationality is the ability to get your priors
> and U right.
> --
> Gordon Worley

Let me see if I can transform this into something I like ;)

Let's assume we have a system at hand with certain goal, embedded in a
certain environment, and with a processor of a certain type and capability

Then, you can assess the effectiveness of the system *by the standards of
the assumed universe*, by assessing

-- how well its actions achieve its goals

-- how well an optimally intelligent system could have achieved those same
goals given the same processor type/capability and the same basic
perception/action domain

The "effectiveness ratio" A/B is an index of how well the system has
succeeded, if the ratio is 1 then the system is perfect, if the ratio is 0
then it's a total failure.

I don't call this intelligence or rationality, i call it effectiveness,
because I've made no requirement that the goal be complex, and I consider
intelligence as the ability to achieve complex goals in complex

NEXT, you're assuming that the behavior of any system can be modeled by
assuming the system is carrying out *approximate* probabilistic inference
with certain parameters. In other words, any system can be matched with an
"inferential system model."

Then, you're saying that those systems with higher effectiveness ratios,
given a certain processor type/capability and a certain perception/action
domain type, are going to tend to correspond to inferential system models
with certain particular parameter sets.

-- Ben G

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