Analyzing the Influence of Instant Messaging on User Relationship Estimation

Recent years have seen a plethora of new mobile social networking services, given the widespread and ubiquitous availability of smartphones. However, from a user's perspective, two fundamental problems underlie these services: Undesired interruptions and privacy violations. Understanding user relationships can help smartphones to provide appropriate decision support for improved notification management and content sharing, thus mitigating these negative effects. In this work, we investigate the influence of mobile instant messaging (IM) services in estimating the type and strength of user relationships. To this end, we implemented an Android-based application to gather users' historical communication data and ran a study to collect manual assessments for each smartphone contact. Our user study shows that friends and hobby-related contacts tend to communicate more using IM services, whereas family and work-related contacts tend to use calls. Furthermore, our machine learning models estimate the social circles with an average accuracy of 77%, and distinguish between strong and weak relationships with an average accuracy of 76%, therein.