On Predicting User Intent

Book One

Volume

Proceedings of the AAAI Conference on Artificial Intelligence, 20

Student Abstracts

Recent advances in computing capability and approaches have underlined the need for user support in task execution. This is especially evident in dynamic, real-time systems where decision making is critical and errors are costly. Environments such as air traffic control centers leave little time for human operators to make critical decisions. The main outcome of working in such an environment is cognitive overload, which leads to a significant error rate in decision making. Many in the human-computer interaction community have concentrated on looking for a solution to relieving cognitive overload through the application of user interface engineering conventions. These methods are expressed as rules which, when followed, are said to result in user interfaces that have high usability. However, regardless of the user interface, the operator must still make decisions and perform all of the tasks himself. This paper presents an outline for creating an intelligent agent to assist the user. The role of the agent is to attempt to predict the user’s intent. The agent can then act on its perceptions with the goal of accelerating the task’s completion and easing the cognitive load of the user by manipulating the user interface.