Widening the Role of Group Recommender Systems with CAJO
Francesco Ricci, Amra Delić
TL;DR
The paper addresses the lag in real-world adoption of Group Recommender Systems and argues that effective group decision making has been under-targeted. It introduces CAJO, a four-role framework (Coach, Arbiter, Judge, Oracle) to enable structured human–AI collaboration across the group decision process. A survey of GRS state-of-the-art covers aggregation methods, single-shot and conversational systems, and notable platforms (ARCADES, STSGroup, CHARM, MUCA), highlighting data gaps and evaluation challenges. It calls for a paradigm shift to holistic, process-aware group decision support and outlines multidisciplinary research directions and societal applications to realize fairer, more productive group decisions.
Abstract
Group Recommender Systems (GRSs) have been studied and developed for more than twenty years. However, their application and usage has not grown. They can even be labeled as failures, if compared to the very successful and common recommender systems (RSs) used on all the major ecommerce and social platforms. As a result, the RSs that we all use now, are only targeted for individual users, aiming at choosing an item exclusively for themselves; no choice support is provided to groups trying to select a service, a product, an experience, a person, serving equally well all the group members. In this opinion article we discuss why the success of group recommender systems is lagging and we propose a research program unfolding on the analysis and development of new forms of collaboration between humans and intelligent systems. We define a set of roles, named CAJO, that GRSs should play in order to become more useful tools for group decision making.