lansky@VENICE.AI.SRI.COM (Amy Lansky) (11/06/87)
EVOLVING KNOWLEDGE AND TMS
Anand S. Rao (ANAND@IBM.COM)
IBM T.J. Watson Research Center and Sydney University
(joint work with
Normal Y. Foo
IBM Systems Research Education Center and Sydney University)
11:00 AM, MONDAY, November 9
SRI International, Building E, Room EJ228
The traditional view of knowledge in the AI literature has been that
'Knowledge' is 'true belief'. The semantic account of this notion
suffers from a major problem called Logical Omniscience, where the
agent knows all valid formulas and his knowledge is closed under
implication. In this talk we propose an alternative viewpoint where
knowledge or EVOLVING KNOWLEDGE (as we call it) is treated as
'indefeasibly justified true belief'. This notion of knowledge solves
the problem of logical omniscience and also captures the
resource-bounded reasoning of agents in a natural way. We give the
semantics and axiomatization of this logic of evolving knowledge and
discuss its properties.
The logic of evolving knowledge also serves as the logical foundation
for the Truth Maintenance System (TMS). We provide a transformation to
and from TMS nodes to formulas in this logic. We show that a set of
nodes has a 'well founded labelling' iff their corresponding IN nodes
are 'satisfiable' in this logic and their corresponding OUT nodes are
'not satisfiable' in this logic. We conclude the talk by comparing our
logic with Autoepistemic Logic, Deduction model of Belief and the
Awareness model of belief.
VISITORS: Please arrive 5 minutes early so that you can be escorted up
from the E-building receptionist's desk. Thanks!