Abstract
Generating a sequence of music tracks recommendations to a group of users can be addressed by balancing the users' satisfaction for a set of recommendations (the playlist), rather than finding items that individually provide good average satisfaction to the users. In this paper we introduce a 'Balancing' technique that builds a tracks' sequence iteratively while constantly balancing users' satisfaction levels. In a live user study we have shown that it produces playlist recommendations that are better than those generated by the average preference aggregation method and comparable to those manually compiled by the group members.