Group decision making with collaborative-filtering 'in the loop': Interaction-based preference and trust elicitation

Recommender systems have been one of the main methods to overcome the information overload problem in many domains including tourism. A solo traveller may easily create an itinerary according to his or her taste. However, when it comes to a group of people, it is often difficult to find the most suitable places to please everyone's preferences, not only because of lack of prior information about a place but also owing to the difficulty to take influence among group members into account. In our work, we introduce a model for Group Decision Making that uses YouTube API to gather rich video contents associated with a finite set of alternatives, e.g. travel destinations. Instead of eliciting subjective opinions on alternatives directly, preferences are built upon a Collaborative Filtering approach, based on each participant's watch history and interaction with items. Randomly chosen videos from different group members are also recommended to each target user in order to infer trust information within the group. We then use trust information to obtain an aggregated group preference for determining places to visit. An application example based on YouTube API shows that a higher degree of interaction consolidates a trust network, resulting in informed decision results.