Figure 1.
An in-game screen shot from AION, a popular MMORPG.
AION’s design puts heavy emphasis on cooperation for success in gaming, generating rich and detailed data of collaborative human interactions. Copyright NCSoft, Inc.
Figure 2.
The definition and the graphic representation of the six AION networks.
The Realm-vs-Realm design of AION where gamers belong to one of two tribes (Heavenly and Diabolical) that cannot communicate is evident in the existence of two similarly-sized large components. Red-colored nodes indicate exceptionally high-degree nodes.
Table 1.
Basic network characteristics of AION interaction networks and other popular social networks.
Figure 3.
The Pearson correlation coefficient between the in- and out-degrees of nodes (solid), and the reciprocity of edges in the AION interaction networks.
The four social-type networks (F, PM, PI, and T) can be further divided according to the reciprocity, the low value of which in Party Invitation and Trade interactions are believed to indicate significant strategizing in the latter two cases.
Figure 4.
The in- and out-degree distributions in AION are most dissimilar in the case of the Mail and Shop networks.
In Shop, the highly skewed in-degree distribution is caused by the existence of œmagnate merchants In Mail, the outliers in the out-degree nodes specify a special class of high-volume senders such as in-game managers.
Figure 5.
Network motif analysis of node triplets reveal detailed interactions patterns in directed networks.
(a) The thirteen possible motifs composed of three nodes in a directed network. (b) The fractions of each motif type in each of the six networks. Motifs accounting for fewer than 10% of the motifs are not shown. Friendship, Mail, and Shop each show one dominant motif, consistent with the high or low reciprocity found in the networks. (c) A closer look at the (normalized) Z-score triad census of Private Messaging, Party Invitation, and Trade networks where no dominant motif is evident, we used the Z-score method is employed to determine significantly over- and underrepresented triangular motifs. Overrepresented motifs (with ) are indeed closed triangles, consistent with the relatively high clustering tendencies in these networks.
Figure 6.
Pairwise network overlaps indicating the similarity or dependence between interactions.
(a) The link overlap. The largest link overlap is found between the Private Messaging and the Shop networks, reflecting the fact that private messaging (for bargaining) nearly always precedes the sales of items via the Shop interaction. (b) The node overlap quantifying the node degree overlap between different networks. High degree overlaps occur between the four social-type networks, indicating that many gamers make a fair mix of the actions. The low degree overlaps in different pairs indicate the role specialization discussed in the text.