Rather than build long-term models, utility-based recommenders match items to
the current needs of the users, taking into account their general tendencies and pref-
erences. For instance, a user may be looking for a particular book, and it is known
from past behavior that old hardback editions are preferred even if it takes longer
to ship them. As is the case with content-based filtering, items can be described in
the system by their features and, more specifically, the utility associated with each
of those features. Aggregation can then be performed as it is with content-based
filtering, although the user profiles and system information may differ.