Social Boundaries in Individual-Group Dynamics: A Multilevel Network Study of a MMOG
Jingyi SunComputer-mediated communication (CMC) is a double-edged sword for social boundaries, as participants of online communities both build interpersonal relations with the recognition of idiosyncratic qualities and build groups based on similar interests. It remains to be answered, however, how such boundaries are established, negotiated and eroded. The neostructuralist theory posits a framework of multilevel agency which suggests norms and hierarchies emerge from a bottom-up process and a top-down process that are intertwined and coevolving, that is, individual members could collectively create new practices that challenge existing relational structures, or collective norms at a higher level are imposed on individuals as opportunities or constraints (Lazega, 2013, 2016).
This study seeks to understand multilevel social mechanisms in a massive multiplayer online game (MMOG) called World of Tanks. The infrastructure of MMOGs usually provokes competition and collaboration for resources and motivates players to form interpersonal networks and join in-game organizations. Data were retrieved from the game publisher’s data warehouse for the North America server from September 2016 to April 2019. There were 18,585 persistently active players affiliated to 5,950 groups. Defining affiliation tie as the median of time duration and selecting groups having more than the median number of members, the data was subset to 8722 players affiliated to 771 groups. A stratified sample of 50 groups were randomly selected based on group age, which yielded 165 players including 9 boundary spanners.
The multilevel network model is organized as follows: the co-playing network among players at the micro-level (network A), player-group affiliation network at the meso level (network X), and textual similarity network of group profiles at the macro-level (network B). The co-playing network was aggregated for thirty-two months and dichotomized on the median of tie weights. In order to attract like-minded players, groups display short paragraphs on their game-linked official webpages to describe who they are and who they want to recruit, thus popularity in the group profile similarity network indicates conventionality in self-identification. Multilevel ERGM analysis was run by MPnet (Wang et al., 2014).
The first key structural mechanism is cross-level closure, that is, whether the boundaries of group affiliation constrain micro-level and macro-level network formation (Brennecke & Rank, 2016; Zappa & Lomi, 2016). In the converged model, ATXAX was positively significant, suggesting individuals sharing the same groups were more likely to play together. In other words, group affiliation did create opportunities and set boundaries for interpersonal communication. ATXBX was not significant, suggesting individuals did not span boundaries based on the similarity of groups, and some individuals might choose to span different groups for new experience. The second important mechanism is the transfer of degree-based hierarchy across levels (Hollway & Koskinen, 2016). StarAXAB, or the transfer of within-level popularity and meso-level popularity, was positively significant, showing conventional groups had more members. ASAXASB, or cross-level assortativity, was negatively significant, showing popular individuals were less likely to join conventional groups. These results show that groups using socially-confirmed language for self-description are viewed as more legitimate in general but not appealing to popular individuals.