AN INTEGRATED FRAMEWORK FOR KNOWLEDGE REPRESENTATION AND ACQUISITION

MANY DIVERSE METHODS OF KNOWLEDGE REPRESENTATION HAVE BEEN USED BY AI RESEARCHERS. NONE APPEAR POWERFUL ENOUGH TO CAPTURE THE DIVERSE TYPES OF KNOWLEDGE THAT PEOPLE POSSESS NOR DO THEY APPEAR FLEXIBLE ENOUGH TO SUPPORT THE ACQUISITION OF KNOWLEDGE. THE PROPOSED PROJECT WILL DEVELOP A KNOWLEDGE REPRESENTATION SCHEME THAT INTEGRATES THREE MAJOR FRAMEWORKS: SCHEMAS/FRAMES, PRODUCTION RULES AND MENTAL MODELS. IT IS ARGUED THAT THESE INDIVIDUAL FRAMEWORKS CAN PERFORM COMPLEMENTARY FUNCTIONS AS PART OF AN INTEGRATED FRAMEWORK: SCHEMAS/FRAMESARE MOST USEFUL AT CAPTURING THE GENERAL STRUCTURE OF KNOWLEDGE, PRODUCTION RULES BEST CAPTURE SITUATION-SPECIFIC PROCEDURES AND ACTIONS AND MENTAL MODELS SERVE AS MECHANISMS WHICH CAN EXPLAIN THE INTERRELATIONSHIPS OF GOALS, PLANS, AND ACTIONS. IT IS HYPOTHESIZED THAT THIS INTEGRATED REPRESENTATION FRAMEWORK CAN SUPPORT KNOWLEDGE ACQUISITION PRIMARILY THROUGH THE USE OF ABSTRACT KNOWLEDGE WHICH REPRESENTS GENERALIZED EXPECTANCIES OF A DOMAIN AND MENTAL MODELS WHICH LINK THAT KNOWLEDGE TO THE SPECIFIC SITUATION. THIS HYPOTHESIS WILL BE TESTED ON A SMALL SCALE AI SYSTEM. THE PROJECT TEAM CONTRIBUTES EXTENSIVE EXPERIENCE IN AI (LEDDO, ABELSON, COHEN, CHINNIS, THOMPSON, AND MCINTYRE) AND EXTENSIVE EXPERIENCE IN PSYCHOLOGY AND THEORIES OF KNOWLEDGE REPRESENTATION (LEDDO, ABELSON, COHEN, AND CHINNIS).