Static Preference Models for Options with Dynamic Extent Thomas Bauereiß, Stefan Mandl, and Bernd Ludwig Dept. of Computer Science 8 (Artificial Intelligence), Friedrich-Alexander-Universit¨ at Erlangen-N¨ urnberg, Haberstraße 2, D-91058 Erlangen, Germany thomas@bauereiss.name, {Stefan.Mandl, Bernd.Ludwig}@cs.fau.de WWW home page: http://www8.informatik.uni-erlangen.de Abstract. Models of user preferences are an important resource to im- prove the user experience of recommender systems. Using user feedback static preference models can be adapted over time. Still, if the options to choose from themselves have temporal extent, dynamic preferences have to be taken into account even when answering a single query. In this paper we propose that static preference models could be used in such situations by identifying an appropriate set of features. 1 Introduction Preferences are an important component of personalized information retrieval and recommendation systems (see [4]). Given a recommendation or information retrieval system that interacts with the user such that the user enters queries and the system produces a list of options the user may choose from, the order of the presented items typically is determined by the quality of the item with respect to some objective value function. Given that this value function is user-pareto- optimal in the sense, that it cannot be changed to improve from the perspective of one potential user without worsening from the perspective of another potential user, any further improvement of the system can only be achieved by personalized user specific adaptations, hence user models and user preferences. In this paper, after discussing the use of standard machine learning tech- niques for classification and regression to represent user preferences in a standard scenario—TV program recommendations—, we focus on a tour recommender. The major difference between tour planning and TV programs is that under the usual granularity of discourse, TV programs typically are considered as singular events (though they have temporal extent) while tours and trips are considered multi-part events. The goal of this paper is to empirically identify proper features of multi-part options in the tour-planning scenario such that standard models of preferences can be used to enhance user experience. Section 2 gives a short account on modeling user preferences: Section 2.1 gives a formal definition of preferences, Section 2.2 contains some notes on pref- erence elicitation, Section 2.3 presents various concrete preferences models, and