Abstract

Explainable AI (XAI) and conversational recommender systems (CRS) have both seen a rise in recent years. However, explanations for CRS (XCRS) are still rare and limited. The use of a design framework for XAI revealed that result, context and counterfactual explanation are of particular importance for XCRS. The importance of counterfactual explanations, especially for end-users with no AI knowledge, is already identified in the field of XAI. However, to the best of my knowledge, this will be the first CRS that integrates counterfactual explanations. In this thesis, a Conversational Movie Recommender will be equipped with a counterfactual explanation system, that gets a suggestion from the user and answers why that film was not recommended and how the user preferences must be changed to get the film as a recommendation. Additionally a context explanation, in which the believed user preferences get stated as explanation and a result explanations that closer explains facts of the recommended movie will be implemented.