Abstract

As a system acquires more knowledge about a task it should be able to take a more active role in learning additional knowledge. To do so, it must be able to re ect upon the knowledge that it has| at each moment| and understand what every piece of knowledge, old and new, contributes to the task. This is the approach taken in expect, a system that represents and manipulates many di erent types of knowledge distinctly and explicitly, and is able to determine what additional knowledge it needs to learn from a user based on its current knowledge about the task. This kind of active participation in the learning process is essential in learning complex tasks, not only from users but also from any other external sources of knowledge, be it information servers or other agents.

Date

January 1, 1970

Authors

Yolanda Gil

Journal

WS-IMLC95