Interfaces that learn : a learning apprentice for calendar management

It is well accepted that knowledge base development and maintenance costs are key obstacles to the further proliferation of knowledge-based systems. We consider here an approach to dramatically reduce this cost by developing learning apprentice systems: knowledge-based advisors which learn from their users throughout their life-cycle. In particular, we present a learning apprentice for calendar management, which allows users to schedule meetings and provides advice regarding parameters such as the meeting time, duration, topic, and location. Each observed user's decision is used as a training example of the correct decision in the current context. The system learns to provide increasingly competent advice by generalizing from these training examples. We present preliminary results showing that knowledge automatically acquired by two learning methods (ID3 and Back-Propagation) compares favorably to manually developed rules when used to schedule meetings for a university faculty member. The system has recently been put into use on a regular basis by one secretary in our environement and is currently undergoing further testing and development. This research is sponsored by the Avionics Laboratory, Wright Research and Development Center, Aeronautical Systems Division (AFSC), U.S. Air Force, Wright-Patterson AFB, Ohio 45433-6543 under Contract F33615-90-C-1465, ARPA Order No. 7597 and by a grant from Digital Equipment Corporation. The view and conclusions contained in this document are those of the authors and should not be interpreted as representing official policies, either expressed or implied, of DARPA, Digital Equipment Corporation, or the U.S. goverment.