The NEO-PEACTIDM MODEL
The Neo-PEACTIDM approach to organizing agent knowledge within the PSCM is
based upon our interpretation of the PEACTIDM model laid out in section
5.3.2 of Allen Newell's book Unified Theories of Cognition
(hereafter UTC). PEACTIDM is a model for immediate behavior which provides
one hypothesis for a way to model extensive interaction with the outside
world. As conceived in Newell's book, it was an abstraction useful for
exploring expertise, i.e. task performance when all behavior occurs in the
Top problem space, or Base-level Problem Space without
impasse. Neo-PEACTIDM is an extension of this view that tries to bring the
rest of the mechanisms of Soar -- specifically impasses and chunking -- into
the PEACTIDM framework. Our primary reasons for advocating this method of
agent construction are that models constructed in this way appear to work
well with chunking, and account for certain timing data.
When using Neo-PEACTIDM, all operators that are selected in the top space
must fit into one of the operator categories outlined in the UTC.
Specifically, there are four kinds of top-space operators, each of which has
a specific, reasonably well-delineated set of functions:
To see documentation of a system built in the Neo-PEACTIDM style, click
here.
This page written by Jill Fain Lehman (jef@cs.cmu.edu)
and Greg Nelson (ghn@cs.cmu.edu)