Knowledge Representation for Self-Aware Computer Systems

1
Knowledge Representation forSelf-Aware Computer
Systems

• Stuart C. Shapiro
• Department of Computer Science and Engineering,
• and Center for Cognitive Science
• University at Buffalo, The State University of
  New York
• 201 Bell Hall, Buffalo, NY 14260-2000
• shapiro_at_cse.buffalo.edu
• http//www.cse.buffalo.edu/shapiro/

2
What Must a KR Have?

• A term to represent itself.
• For beliefs about beliefs
• terms for beliefs
• reified beliefs.
• Term for self not inherently indexical
• Else error of
• Knows(a, P(self)) gt P(self)
• So, extra-logical I register

3
What Must a KR Have?

• Terms for acts
• For memory of past acts
• Model of time
• Extra-logical NOW register
• To distinguish Im doing from I did
• Way to distinguish durative
• from punctual acts.

4
Existing KR Languages

• FOL Sufficient
• Issue is domain design of functional terms.
• E.g. reified agents, propositions, acts, times,
• SNePS
• Modal logics not needed if do above.
• Non-monotonic logics Independent issues
• Unless can have self-awareness
• without memory of past acts
• and dont want a model of time.

5
Place of Mind-BodyConnections

• Need both
• mind to body
• and body to mind
• for intelligent systems.
• SNePS has both.
• But its body to mind that gives self-awareness.

6
Research Issues

• Not what a KR language can represent,
• but where the beliefs come from.
• Architecture s.t. the mind is aware of the body.
• I.e. that inserts beliefs into the mind
• from the body.
• Awareness of reasoning?
• Reasoning is acting!