Like designers of recommender systems for individuals, those who design recommender systems for groups can benefit greatly from a thorough understanding of human decision making and ways of supporting itin particular decision making that occurs in a group context. But a comprehensive analysis with regard to groups has so far been lacking in the recommender systems field. This talk will present such an analysis in an accessible way, referring workshop participants to a recently completed chapter for the Recommender Systems Handbook (3rd edition) for further details and references [1] (a preprint is available on request from the first author).

Relevant knowledge about decision making in groups can be found in research on group decision making, group dynamics more generally, negotiation, and group decision support systems as well as in the practical experience of human group facilitators. To see the relevance of these areas, we will look at seven widely accepted principles derived from them that have important implications for the design of group recommender systems.

We will then consider two high-level approaches to using this type of knowledge in a recommender systems context: supporting interaction among group members vs. predicting the results of interaction without allowing the interaction to occur. The approaches differ greatly in their applicability to particular recommendation scenarios.

We will discuss concrete examples of how group recommender systems have appliedor could applythese ideas and results. These examples are organized in terms of an extension to groups of the previously published Aspect and Arcades models, which distinguish the diverse ways in which people make choices and in which their choice processes can be supported.