Evaluating different preference elicitation methods for a mobile recommender system in a field trial
Recommender systems offer product advice or information users might be interested in. In the tourism sector information needs are directed towards sights and restaurants. Today’s agent technology provides the opportunity for intelligent tour recommenders to organise a personal walking tour and to provide user specific guidance and interpretation to create an individualized experience. This is the main objective of the Dynamic Tour Guide (DTG, ) - a mobile agent that elicits personal generic preferences, selects attractions, plans individual tours, provides navigational guidance and offers location based interpretation via a mobile device. But intelligent recommendation deliveries require knowledge about the user’s interests. A field trial served to evaluate methods for the elicitation of generic preferences and to reason about their general necessity by defining the following research targets: Is it possible to seed generic interest profiles that allow the accurate prediction of rankings of concrete sights? And are these profiles sufficiently diverse to make personalised tours based on individual interests a real improvement towards standard tours? The results suggest that ranking of simple categories constitutes the most effective means of capturing user preferences using a mobile device and that standard tours would not adequately represent the wide array of interest profiles identified.