Social Architecture and the Emergence of Power Laws in Online Social Games

Online social games, according to this analysis, exhibit a consistent power-law pattern in which a small fraction of players generates a disproportionate share of social interaction. The work proposes that the scaling exponent α is a constant, comparable measure of a game's sociability, and that game mechanics deterministically shape this macroscopic behavior largely independent of the particular player population. The central thesis is that the combination of mechanics, interfaces, and social context, termed "social architecture," can be characterized quantitatively through this exponent.

The evidence rests on analysis of six game datasets, including Fighters' Club, Familiars 1, and PASION Fruit, where the power-law pattern appeared consistently. A particularly telling comparison pairs PASION Fruit with Fruit Loot, which share identical mechanics; their exponents came out nearly identical at approximately 0.77, supporting the mechanics-driven interpretation. Reported values differ widely across titles, with PASION Fruit at about 0.773 and Fighters' Club at about 2.032, indicating that α can distinguish architectures from one another.

Beyond comparison, the exponent serves as a diagnostic tool. In the case of Hugged, a 15-notification API limit produced multi-scaling behavior, with α near 0.97 below the threshold of k=15 and near 4.16 above it. The author interprets this break as a usability bottleneck, attributing it to an increased cost of attachment among players, and presents it as evidence that scaling analysis can reveal architectural problems that are otherwise difficult to observe.

The analysis draws its theoretical grounding from network science, including scale-free and small-world network research, and from studies on social network size. Its scope encompasses online social games, with empirical focus on the specific titles examined, while the closing material supplies the funding context and supporting literature rather than independent findings.