Does diversity kill OSNs?
Laszlo Lorincz, Julia Koltai, Johannes Wachs, Karoly TakacsPeople have contact with diverse groups of others. Family members, old school friends, colleagues, and neighbours can be very different with regard to their interests, attitudes, or political preferences. Opinions and norms therefore, that are acceptable in one group can be condemned in another. In the offline world, these differences rarely conflict, as members of the different groups seldom interact with each other. On online social networks (OSNs), however, these audiences get mixed more often, which creates the tension associated with context collapse.
In this paper, we analyse the role of context collapse in the abandonment of iWIW. iWIW was the major Hungarian pre-Facebook OSN, founded in 2002. At its peak, had more than 3.5 million users in a country with a population of 10 million, where this meant two-third of all Internet users at that time. After 2008, with the appearance of Facebook, it started to lose its popularity and after the continuous decline in user activity, iWiW was finally shut down in 2014.
Earlier studies explored the mechanisms of the collapse of iWIW from different perspectives; by explaining it with cascading mechanisms, and by analysing the social capital differences of its users. The role of context collapse, however, has not yet been studied on this network, and we do not know examples of large-scale quantitative studies on the role of context collapse in the decline of other OSNs either.
Based on literature about context collapse, we expected that the mechanism contributes to abandoning the site early.
For the analysis, we selected a random sample of 10,000 users, and observed their ego-networks between 2007 and 2012 on a monthly basis. We defined context collapse with the following conditions:
1. A user has several circles of friends, which only hardly overlap, and
2. these circles of friends have distant social background.
Louvain community-detection algorithm was used to identify these social circles. We measured the distinctness of these social circles by the modularity of the ego-networks, and the social distances between the circles by their average differences in age, gender, and location.
We predicted the abandonment of the site (when a user becomes inactive) by discrete time survival models. We found that contrary to our expectations, high modularity of the ego-networks did not facilitate, but retracted the users exiting the site, similarly to having socially more distant circles. These effects are robust across different specifications (parametric and non-parametric hazard functions, alternative definition for exit using the time of making the last connection, and separating the early and later periods of the OSN’s life cycle).
Based on Boyd (2014), we interpret these results as that those, who perceive context collapse might have a strategy of using different OSNs the same time. In these cases, different OSNs are used for different communities. These users go to new platforms relatively early, but also stay on the old ones for a longer period.