Such automated recommendation-based techniques can be helpful in social media systems such as email to choose with whom to share a message. However, these techniques discourage users from creating groups for controlling information flow on public and large social networks. These suggestion-based approaches still put a relatively high burden on users to verify friend suggestions one at a time. If one user sends ten messages on an OSN, this requires verifying all of the recipients for all ten messages and may become a hinderance for frequent users of OSNs.
Using automated approaches and allowing for minor user modification to create groups in OSNs is an alternative to the existing recommendation-based grouping techniques. This approach creates fully populated groups from the onset and then allows the user to modify them. One method for creating such groups uses clustering algorithms to automatically detect groups in OSNs. While the feasibility of using clustering algorithms for group creation in OSNs has been investigated before, less is known about the benefits and drawbacks of using such automated friend grouping approach within a social media interface.
In this work, we present a grouping tool that automatically creates groups within Facebook using three different algorithmic techniques. This tool creates groups from a Facebook friendship network and then allows for human modification of groups. We conducted a study in which we asked participants to work with our tool and modify their populated friend groups as they wanted. During 18 semi-structured interviews, we investigated the advantages and disadvantages of automated friend grouping in OSNs. Mainly, we found a significant automation bias which meant different algorithms for grouping friends affected the final groups one person created.
Eslami, M., Aleyasen, A., Zilouchian Moghaddam, R., Karahalios, K. Evaluation of Automated Friend Grouping in Online Social Networks. CHI EA 2014. pdf
Eslami, M., Aleyasen, A., Karahalios, K. Make Community Detection More Human. HCIC 2013. pdf