Empowering recommender systems using trust and argumentation

Recommender systems (RSs) use the opinions of members of a community to help individuals in that community identify the information most likely to be interesting to them or relevant to their needs. These systems use the similarity between the users and recommenders or between the items to form recommendation list for the user. We believe that, various interactions and arguments exchanged in favor or against are responsible for the eventual result of a recommendation process. Therefore, besides recommendations it is vital to determine the users' response on such interactions to determine more accurate trust estimates for users in the system. Hence, this paper proposes a novel fuzzy and argumentation based trust model which is also integrated within the practical reasoning of agents in the multi-agent recommender systems. This integration allows the agent to take trustworthy decisions and reason about them as well. The user is also able to make a wiser selection in case there are conflicting opinions related to a specific product or the user comes across a new, unseen product and is indecisive about it. As a result it improves recommender's persuasive power and user's trust in the system resulting in an increase in the user's acceptance of the recommendations. The experiments performed with a Book Recommender System (using a hybrid recommendation approach), confirms that the variant implemented with the proposed approach performs better than those using conventional methods. Results obtained from evaluation metrics showed that the recommendations were more accurate, relevant and novel.