Abstract:

Preference elicitation is a well-known bottleneck in decision analysis and
decision automation tasks. The best description of a user's preference is
via a utility function. However, obtaining a good utility function from a lay
user requires the assistance of an expert. For example, suppose that you
want to help a customer of an online merchant to choose an appropriate
configuration for a PC, or select the vacation that is most suitable for
him. An expert decision analyst will not be around. This is why we have been
developing Conditional Preference Networks (CP-nets), a family of formally
sound qualitative models for representation and reasoning about preference
that are based on intuitive statements which lay users find natural.  In
this talk I will present the CP-nets model, describe its computational
properties and various application-related issues, and discuss some current 
and future work in this area.

* Joint work with Ronen Brafman (Ben-Gurion University).