Much is written about the benefits of computer-supported collaborative learning (CSCL) but many accounts report problematic interactions and less than desirable outcomes for attempts at CSCL. A theoretical approach is presented that is grounded in a social identity perspective of groups and seeks to promote and support successful collaborations. Using this approach, along with examples from laboratory and field studies, key concepts such as cohesion, participation, accountability and group norms are addressed and contrasted with more traditional approaches. It is suggested that design features that ‘collectivise’ rather than ‘personalise’ CSCL can promote psychological issues that are crucial for successful collaborations (i.e. group cohesion, accountability to the group, and increased participation), and norms for success can be developed through intergroup comparisons. The theoretical approach is also presented as a platform from which further investigations and predictions can be made.