Proactive Recommendation Delivery

Abstract

The main purpose of Recommender Systems is to minimize the effects of information/choice overload. Recommendations are usually prepared based on the estimation of what would be useful or interesting to users. Thus, it is important that they are relevant to users, whether to their information needs, current activity or emotional state. This requires deep understanding of users' context but also the knowledge of the history of previous users' interactions within the system (e.g. clicks, views, etc.). But even when the recommendations are highly relevant, their delivery to users can be very problematic. Many existing systems require active user participation (explicit interaction with the recommender system) and attention. Or, on other side of spectrum, there are RS that handle recommendation delivery without any consideration for users' preferences of when, where or how the recommendations are being delivered. Proactive Recommender Systems promise a more autonomous approach for recommendation delivery, by anticipating information needs in advance and acting on users' behalf with minimal efforts and without disturbance. This paper describes our work and interest in identifying and analyzing the factors that can influence acceptance and use of proactively delivered recommendations.