With social media being a big force in modern virtual and non-virtual relationships, there are some interesting analyses done on the signed networks that exist within the platforms. In the article titled Signed Networks in Social Media, the three researchers Jure Leskovec, Daniel Huttenlocher, and Jon Kleinberg, take a look at the network representations of 3 online platforms Epinions, Slashdot, and Wikipedia. Epinions, an ecommerce platform, Slashdot, an online social news website, and Wikipedia, the online encyclopedia all incorporate aspects of a positive/negative relationship for their users.
On Epinions, users can give ratings not only to the items sold on the marketplace, but also to other users on the site as well. On Slashdot, users can label other users as ‘friends’ or ‘foes’. Wikipedia periodically has volunteer candidates considered for a promotion into an admin role on the platform, and the community casts public votes for whether they are for or against this promotion.
The table below shows the data for the signed relations within these networks. These figures show that for weak structural balanced networks, triangles with two positive edges are massively underrepresented. Triangles with three positive edges are overrepresented when compared to the other types of triangles.
The data from these 3 platforms give further developments for signed networks. For triangles within signed networks, we label balanced triangles as triangles with an odd number of positive edges. Balanced signed network analysis is usually applied to undirected networks. This is where the ideas of “the friend of a friend is a friend”, and “the friend of my enemy is my enemy” came from. An idea that is explored in this paper is the theory of status, which has applications in directed networks. Status in signed networks indicate that the creator of positive links view the recipient of the relationship as having higher status, and the creator of negative links view the recipient of the relationship as having lower status. A good way to distinguish between these two theories is their proposed situation: A links positively to B, B links positively to C, according to the balance theory, A should link positive to C as well as C is a friend of a friend. However, in the status theory, since C > B > A, C should view A as having lower status in the hierarchy.
Since social media and many of these online platforms tend to be used for personal satisfaction, such as finding interests or a community to belong to, it makes sense that positive triangles are overrepresented. People tend to dislike being involved with online conflict, as many internet celebrities nowadays are constantly being very cautious in order to have a good public image. It is also relevant to average users as well, people want to maintain good relations and a good public image to their fellow friends on these platforms. On websites like Linkedin, keeping good connections and avoiding creating bad ones are important, as having more of a positive profile opens up many more job opportunities.
Social media is an entity that has now been decades old, and is built on the relationships of its users. These analyses done on the networks between the relationships on these platforms can teach us the trends between how users use their service, and additional data collected can even further improve the user experiences for everyone around the world.
Source:
One reply on “Signed Networks in Social Media”
Cool job with analyzing signed relationships within social media! The brief commentary on how social media encourages positive triangles got me thinking about a few things regarding the current and future climate of social media. Specifically, it seems that many social media platforms avoid functionality that creates negative relationships. For example, Facebook doesn’t really have a dislike button or a negative “reaction emote” when reacting to a post (for good reasons including avoiding bullying etc.). I’m just wondering if this trend or norm of avoiding negative relationships and only encouraging positive ones will limit the quality/usefulness of network analysis, since all those negative relationships will be gone, which in turn eliminates signed network analysis.