Human dynamics analysis in online collaborative writing

Investigating the human online behavior has become a central issue for understanding human dynamics in recent years. In this paper we analyze the temporal and content-updating statistical properties of online collaborative writing based on Wikipedia data. Online collaborative writing is one of the important and widespread human online behaviors, which is of great apphication. Empirical result shows that the distribution of inter-event time in collaborative writing is on the multi-scale. That is to say, two time intervals that range from 1 min to 30 min and 30 min to 24 h both obey power-law distribution with exponents equal to 1.62 and 1.16 respectively, while the interval larger than 24 h obeys a distribution whose cumulative form is F ( τ )∝ τ － b － a log( τ ) . More investigatons show successive updating behavior and mutual updating behavior working together to lead to the multi-scale distribution of inter-event time. Successive updating behavior leads to the power-law distribution with an exponent 1.62 of interval within 30 min while mutual updating behavior leads to the power-law distribution with an exponent 1.16 of interval ranging from 30 min to 24 h. Furthermore, we find that reverse updating repeats frequently in collaborative writing. The proportions of reversing updating and the updating size are strongly relatively reflect that the updating size is a main reason leading to the relevant content to be preserved. The bigger the updating size, the harder it would be preserved. More statistical analyses imply that "watching dog" and "edit war" exist in Wikipedia editing. Those results are very helpful to deepen the understanding of the human collective behavior, especially of the collaborative developing behavior.