METHODS AND TOOLS FOR ASSESSING LIMITS OF SYSTEM INTELLIGENCE

INTELLIGENT SYSTEMS TECHNOLOGY IS ENVISIONED AS BEING AN IMPORTANT COMPONENT OF FUTURE AVIATION AND SPACE SYSTEMS. EXPERT SYSTEMS FOR MISSION PLANNING, FLIGHT CONTROL, AND FLIGHT MANAGEMENT ARE PRIMARY EXAMPLES. THE ABILITY TO BUILD INTELLIGENT SYSTEMS HAS BEEN DEMONSTRATED. HOWEVER, WE DO NOT UNDERSTAND THEIR PROPERTIES IN WAYS THAT WE UNDERSTAND, FOR EXAMPLE, CONTROL SYSTEMS. LIMITATIONS OF THE PHYSICAL AND BEHAVIORAL MODELS THAT UNDERLIE INTELLIGENTSYSTEMS CAN LEAD TO UNDESIRABLE BEHAVIORS ANALOGOUS TO OSCILLATION AND INSTABILITY IN TRADITIONAL SYSTEMS. THIS PROJECT WILL DEVELOP METHODS AND TOOLS FOR ANALYZING INTELLIGENT SYSTEMS AND PREDICTING LIKELY UNDESIRABLE BEHAVIORS. THE OVERALL METHODOLOGY WILL INVOLVE ITERATIVE USE OF ANALYTICAL AND EMPIRICAL METHODS TO ASSESS INTERNAL CONSISTENCY AND EXTERNAL VALIDITY, RESPECTIVELY. THE EVENTUAL PRODUCT WILL BE COMPUTER-BASED TOOLS FOR ANALYZING DESIGNS OF INTELLIGENT SYSTEMS SUCH AS ARE LIKELY TO BE PARTOF FUTURE NASA SYSTEMS, AS WELL AS A PLETHORA OF COMMERCIAL SYSTEMS.