Eye tracking document authours and web searchers

Sutiable for: PhD or minor thesis.

Background

When evaluating search engines, there are many implicit assumptions around how people actualy process the information with which they are presented. For example, it is known that there is a "trust bias", where people will click on highly ranked results even if they are irrelevant. While gross assesments of interaction with search engines have been performed, little is known about individual variations in interaction, and how this can be captured and used to improve search engines.

In particular, can a "language model" of a human's information need be created that would allow results to be presented in a way that is readily understood by the individual searcher?

Research Questions

• What learning effects take place as a user interacts with a results list?
• Can the results list be dynamically altered to incorporate these learning effects?
• Can a "language model" of a user be formed that allows translation of search results into a more useful form?
• Can studying document authouring give us insights to these problems?
• Can these human factors be incorporated in batch evaluation of search engines?

Student's Background Required

The project would involve programming in either C, Java, Matlab or R.

Some background in psychology would be an advantage.