User Control and Explanations in Recommender Systems : In Search for a Better Recommendation Process

Abstract

Recommender engines have become a highly popular choice for filtering information in digital services. And even though the algorithms used in them have become extremely good at creating accurate recommendations, there still remains problems related to the overall user experience. Such problems include trust issues, irrelevant recommendations, shortage of diverse recommendations, experiencing negative emotions as well as the feeling of lack of control. Two methods that have been offered as solutions to the remaining challenges are increased user control and explanations. User control enables the users to affect the recommendations process to better fit their needs, whereas explanations offer information about it. On these premises, a literary review of 11 articles on recommender systems, user control and explanations was conducted. The objectives for this study were to examine the interaction between the user and the recommender system, and to determine if increased user control and explanations would have a positive effect on the user experience. The study observed that the target of interaction varied between different recommender systems, and that the users wanted to utilize the offered interaction mechanisms, although additional properties were also demanded in some cases. Additionally, traditional user interface components were found to be a popular way for implementing interaction in recommender systems. The findings of the study also indicate that users prefer recommender systems that offer user control and explanation features. Implementing user control and explanations into recommender systems improved the quality of recommendations and enhanced the overall user experience, although some individual aspects of the user experience were negatively affected. Moreover, the level of control and personal characteristics of the user played a role in various aspects of the interaction.