Tie Strength on Social Networks
http://leonidzhukov.net/hse/2011/seminar/papers/chi09-tie-gilbert.pdf
The research paper “Predicting Tie Strength With Social Media” by Eric Gilbert and Karrie Karahalios is trying to answer the following question: Can social media data predict tie strength? The authors present a model that is capable of predicting tie strength in social media platforms. The model was tasted on 2,000 social media ties and it proved to predict the tie strength with 85% accuracy. The authors hope than one day their model will be used to improve the design of current social media sites (e.g., information prioritization, personalized privacy settings, etc.).
To answer their research question Gilbert and Karahalios recruited 35 participants who had to rate the strength of their Facebook friendships (62 rated friendships per participant on average). The researchers also collected data about the participants’ Facebook profiles and friendships, which was later used to run the predictive tie strength model. Each Facebook friendship was assessed for tie strength and described by more than 70 numeric indicators such as wall words exchanged, number of mutual friends, appearances together in photos, etc. To better understand their model limitations, the authors also engaged in follow-up interviews about friendships/ties that they struggled to predict.
This paper is strongly related to the topics covered in class about weak and strong ties. Moreover, the most relevant topics covered in class about tie strength were triadic closure and the Strong Triadic Closure property. Triadic closure is a concept that claims that if two people in a social network have a friend in common, then there is more likelihood that those two persons will become friends. On the other hand, the Strong Triadic Closure property claims that if a person has a strong tie with two different neighbors, then these neighbors must have at least a weak tie between them. Even though triadic closure was not specifically mentioned in the paper, some of the variables that the researchers used to model the tie strength between Facebook friends are strongly related to the concept. For example, Gilbert and Karahalios used the number mutual friends, which is based on the concept of triadic closure, as variable to model tie strength. Furthermore, the Strong Triadic Closure property has a lot of potential applications in social networks. For example, social networks could base their “friend suggestion/recommendation” engines on such property. If a person has a strong tie with two neighbors, social networks should suggest these neighbors to friend each other.