Characteristics of Human Behavior in an Online Society

This study analyzed user behaviors on Epinions and generated the following findings. The frequency of the trusting, writing, and rating behaviors of Epinions users demonstrated a power law distribution. Only a small group of users were active on Epinions then. The majority of users obtained trust from only a small group of users, whereas only a small group of Epinions users were trusted by a sizable group. Therefore, most Epinions users were information receivers, whereas only a small group of users were information providers. The intervals of writing, trusting, and rating behaviors on Epinions followed power law distributions, conforming to the results of previous human dynamics studies. The interval of reciprocation also exhibited a power law distribution. The small proportion of reciprocation among Epinions users reflected a distinction between an online trust-relationship network and a real-world social network. Trusting, writing, and rating behaviors on Epinions exhibited patterns similar to power law declines and their frequency fluctuated in later periods. Finally, this study proposed a model on the basis of diminishing user activity over time. By describing the probability of participating in a behavior on Day t and conducting simulation calculations, this study found that this model fit the power law distribution characteristics of user behavior intervals and could thereby serve as an adequate human dynamics model.