Recommender System

A significant problem of recommender systems, as with artificial intelligence systems in general, is their lack of explainability. If the recommendations are not explained properly to a user, there may be a lack of trust from the user in the system, leading to ineffective feedback and, consequently, an inability of the system to adapt to the user’s preferences effectively.

Our hybrid recommender system uses argumentation technology as an underlying framework for providing explanations of its recommendations to users. These frameworks allow different forms of explanation, e.g. visual or linguistic, to ensure that users understand how the system is deriving its recommendations. The system then elicits and adapts to feedback from users in such a way that future recommendations are guaranteed to be more suited to each user’s individual preferences.

Papers

• Antonio Rago, Oana Cocarascu, Francesca Toni: Argumentation-Based Recommendations: Fantastic Explanations and How to Find Them. IJCAI 2018