Argumentative explanations for interactive recommendations

A significant challenge for recommender systems (RSs), and in fact AI general, is the systematic definition of explanations outputs such a way that both themselves are able to adapt their human users' needs. In this paper we propose an RS hosting vast repertoire explanations, which customisable users content format, thus explanatory requirements, while being reasonably effective (proven empirically). Our built on graphical chassis, allowing extraction argumentation scaffolding, from diverse varied argumentative recommendations can be obtained. These interactive because they questioned by support adaptive feedback mechanisms designed allow self-improve theoretically). Finally, undertake user studies vary characteristics showing general preferences more information, but also tastes diverse, highlighting need our adaptable RS.