CMU Artificial Intelligence Repository
MULTAGNT: Prolog simulation of multi-agent tutoring.
areas/reasonng/meta/multagnt/
MULTAGNT is a simulation of error detection and identification in a
small tutoring setting, consisting of a Teacher and a Learner. The
objective of MULTAGNT is to show, how through meta-logical knowledge the
Teacher can detect and identify deficiencies in the Learner's knowledge.
The meta-logical knowledge used consists of predicates:
- can_do, for determining what a theory of an agent can prove
- cannot_do, for determining what a theory of an agent cannot
prove
- demo_trace_2, for determining the erroneous clause in an
agent's theory
which are used by an agent (usually the Teacher) to judge the
knowledge of another agent (referred-to as Learner). Agent's are
represented by sets of theories, where a theory is a set of clauses.
In this setting the Teacher's knowledge is assumed to be correct.
Thus, it can be used as reference knowledge, for generating question
posed to the Learner and to judge the knowledge of the Learner.
Origin:
src.doc.ic.ac.uk:packages/prolog-pd-software/ (146.169.2.1)
as the file multagnt.zip
Version: 21-AUG-93
CD-ROM: Prime Time Freeware for AI, Issue 1-1
Author(s): Thomas Hoppe, Technical University of Berlin
Yiu Cheung Ho, King's College London.
Keywords:
Authors!Ho, Authors!Hoppe, Meta Reasoning,
Multi-Agent Tutoring, Prolog!Code, Reasoning!Meta Reasoning
References: ?
Last Web update on Mon Feb 13 10:27:50 1995
AI.Repository@cs.cmu.edu