USER INTERACTION WITH SELF-LEARNING SYSTEMS

MILITARY INTEREST IN ARTIFICIAL INTELLIGENCE (AI) AND EXPERT SYSTEMS IS GROWING IN RESPONSE TO NEEDS FOR REDUCTING MANPOWER-INTENSIVE TASKS AND PROVIDING AID TO DECISION MAKERS WHOSE INFORMATIONPROCESSING CAPABILILTIES ARE BEING OVERTAXED. THE CUTTING EDGE OF RESEARCH ON EXPERT SYSTEMS CONCERNS SELF-LEARNING SYSTEMS, NAMELY, SYSTEMS WHOSE KNOWLEDGE BASES AND PROCESSING LOGIC CAN CHANGE IN RESPONSE TO USER COMMANDS OR EXTERNAL DEMANDS. TO ENSURE EFFECTIVE COOPERATIVE PROBLEM SOLVING, USERS OF ADAPTIVE SYSTEMS MAY HAVE TO BE INFORMED OF CHANGES INTERNAL TO SUCH SYSTEMS, THROUGH A DYNAMIC INTERFACE WHICH IS EASY TO COMPREHEND. IT IS PROPOSED TO INVESTIGATE HOW ADVANCES IN THEORIES OF COGNITIVE SCIENCE AND MENTAL MODELS CAN INCREASE OUR UNDERSTANDING OF USER INFORMATION NEEDS AND GUIDE THE DESIGN OF USER INTERFACES WITH SELF-LEARNING SYSTEMS, IN THE FACE OF THE CHANGING BASES FOR THE SYSTEMS' RESPONSES. THREE TASKS ARE PROPOSED FOR PHASE I: (1) EXTEND THEORIES OF KNOWLEDGEMENT REPRESENTATION AND MENTAL MODELS TO DETERMINE USER INFORMATION NEEDS WHEN WORKING WITH SELF-LEARNING AI SYSTEMS; (2) DEVELOP NEW DESIGN CONCEPTS FOR SELF-LEARNING SYSTEMS, TO INCORPORATE NEEDED USER INTERFACES; (3) DETERMINE THE FEASIBILITY OF DYNAMIC ON-LINE DISPLAYS AND RELATED INTERACTIVE TECHNIQUES EMBODYING THESE CONCEPTS.