Recommender systems are agent-based systems that use stored user preferences to locate and suggest items that will be of interest to associated users. These systems will be useful and effective to the extent that they can make meaningful and consistent tradeoffs between conflicting user preferences. Typical application domains for recommender systems (RSs) include recommendations for music CDs and cassettes, movies, books, etc. In any one such domain items to be recommended are selected by a personal assistant agents (PAAs) based on stored user preferences about several domain features. Typically a domain has several features (we will refer to these as domain dimensions). Each dimension consists of a collection of elements, and the preferences of a user is given by his/her ratings of those values on some ordinM or cardinal scales. For example, in an RS for movies, one dimension may describe the type of a movie, and contain elements like horror, comedy, tragedy, musical, action, etc. The preference of a user for different types of movies can be represented by assigning values in the range [0,1] to each of these elements, e.g., a user who really likes musicals will assign a rating close to 1 for musicals. To obtain a recommendation rating for a given item, a RS typically will consider the feature values of that item (e.g., given a movie, the RS selects the movie type, the names of director, leading actor, actress, year of release, etc.), obtains ratings for these values from corresponding dimensions, and then combines these ratings by some evaluation scheme.