Next: Specializing to User Up: Experimental Results Previous: Questionnaires

Tests with simulated users

A learning filtering system must specialize to current user interests, adapt as they change over time and explore newer domains prospecting for potentially relevant information. In this section, the performance of Newt is evaluated for its ability to specialize, adapt and explore. A number of experiments have been performed using simulated user input under controlled situations. This approach allows us to focus on and study the behavior of each sub-component of the news filtering system which would otherwise be impossible in an unconstrained environment. The experiments and the results are described in the following. A sample scenario is presented along with each of the experiments as a vehicle for introducing the user context.



MIT Media Lab - Autonomous Agents Group - agentmaster@media.mit.edu