On the Modelling of an Agent's Epistemic State and its Dynamic Changes

Abstract

Given a set of unquantified conditionals considered as default rules
or a set of quantified conditionals such as probabilistic rules, an
agent can build up its internal epistemic state from such a knowledge
base by inductive reasoning techniques. Besides certain (logical) knowledge,
epistemic states are supposed to allow the representation of preferences,
beliefs, assumptions etc. of an intelligent agent. If the agent lives in
a dynamic environment, it has to adapt its epistemic state constantly to
changes in the surrounding world in order to be able to react adequately
to new demands. In this paper, we present a high-level specification of
the Condor system that provides powerful methods and tools for managing
knowledge represented by conditionals and the corresponding epistemic
states of an agent. Thereby, we are able to elaborate and formalize
crucial interdependencies between different aspects of knowledge
representation, knowledge discovery, and belief revision. Moreover,
this specification, using Gurevich's Abstract State Machines, provides
the basis for a stepwise refinement development process of the Condor
system based on the ASM methodology.