Just-In-Time Information Retrieval

This thesis defines Just-In-Time Information Retrieval agents (JITIRs): a class of software agents that proactively present potentially valuable information based on a person's local context in an easily accessible yet non-intrusive manner. The research described experimentally demonstrates that such systems encourage the viewing and use of information that would not otherwise be viewed, by reducing the cognitive effort required to find, evaluate and access information. Experiments and analysis of long-term use provide a deeper understanding of the different ways JITIRs can be valuable: by providing useful or supporting information that is relevant to the current task, by contextualizing the current task in a broader framework, by providing information that is not useful in the current task but leads to the discovery of other information that is useful, and by providing information that is not useful for the current task but is valuable for other reasons. Finally, this research documents heuristics and techniques for the design of JITIRs. These techniques are based on theory and are demonstrated by the field-testing of three complete systems: the Remembrance Agent, Margin Notes, and Jimminy. Specifically, these heuristics are designed to make information accessible with low effort, and yet ignorable should the user wish to concentrate entirely on his primary task. Thesis Supervisor: Prof. Pattie Maes Title: Professor of Media Arts and Sciences This work was supported in part by British Telecom, Merrill Lynch and the Media Lab’s Digital Life consortium. The views and conclusions contained in this document are those of the author and should not be interpreted as representing the policies or views, either expressed or implied, of the sponsoring agencies or any other person or organization mentioned within. Just-In-Time Information Retrieval by Bradley James Rhodes The following people served as readers for this thesis: Reader Walter Bender Senior Research Scientist MIT Media Laboratory Reader Dr. Elizabeth Mynatt Assistant Professor College of Computing Georgia Institute of Technology