We analyze information diffusion using empirical data that tracks online communication around two instances of mass political mobilization that took place in Spain in 2011 and 2012. We also analyze protest-related communications during the year that elapsed between those protests. We compare the global properties of the topological and dynamic networks through which communication took place, as well as local changes in network composition. We show that changes in network structure underlie aggregated differences on how information diffused: an increase in network hierarchy is accompanied by a reduction in the average size of cascades. The increasing hierarchy affects not only the underlying communication topology but also the more dynamic structure of information exchange; the increase is especially noticeable amongst certain categories of nodes (or users). Our findings suggest that the relationship between the structure of networks and their function in diffusing information is not as straightforward as some theoretical models of diffusion in networks imply.
The data we analyze track online communication through Twitter during the protests that emerged in Spain in May 2011 [5,25] and again in May 2012 [26,27]. Media accounts of the events were quick to attribute to Twitter an instrumental role in the spread of calls for action and the coordination of demonstrations; this role has subsequently been spelled out by ethnographic work and interviews with protesters
The analyses done suggest that, compared to 2011, the network in 2012 was significantly larger and with a higher number of highly connected users at its core, but not necessarily as successful in terms of triggering large chains of information diffusion. During the year that separates the two observations, the network changed its composition significantly: many users that were active in 2011 disappear from the 2012 sample; even more users joined the communication exchange in 2012; and of those staying in the network, some changed their structural position. Figure 3 summarizes these differences. About 7% of the users we capture in our sample discontinued their engagement from 2011 to 2012, and about 41% joined anew in 2012; only about 8% of all users captured by the samples appear in the three observation periods. As panel (b) suggests, the network position of these users changes substantially from year to year, especially in the network of explicit protest communication.