Incorporating contextual information in recommender systems using a multidimensional approach
ACM Transactions on Information Systems (TOIS)
| Hi-index | 0.00 |
The goal of our research was to gather evidence about users' requirements and preferences for ITV recommender systems to be applied in future design and development of these systems. At present, there is a gap between the technological developments in recommender systems and the knowledge about how to ensure user acceptance of these services including how to optimise user interaction with this type of systems. This paper presents two studies that we conducted towards bridging this gap. In the first one we followed an ethnographic approach with interviews, virtual ethnography and field observation in order to gather users' requirements concerning recommendation systems. The information collected allowed us to establish a number of specific requirements and situations for which recommendation systems would be perceived as adding value to their day-to-day lives. In the second study we investigated further the requirements previously gathered from the first study using a methodology called scenario-based requirements gathering with a prototype. The main results of the second study were the validation of most of the requirements gathered in the first study plus some guidelines about user preferences over the means to interact with this type of systems. For instance, we gathered evidence about the importance of providing further explanations on the recommended contents and its impact on the users' perception over the value of these systems. In addition, we have proven the existence of individual differences on the types of recommendation and the level of interaction preferred by different type of users.