Online attention dynamics: The triangle framework of theory, big data and simulations

Online collective actions have great impacts to societies, and we focus on the life cycle pattern of coexisting multiple online cases. We propose the analysis framework of “Theory-Big Data-simulation” (TDS). The theory includes two essential action rules of netizens. We combined macro-level analysis (big data analysis) and micro-level behavior (agent-based modeling), to surpport the theory. For the big data, macro-level life cycle pattern (persistent) is caused by micro-level individual actions. For macro-level verification, basic assumption is that netizens are allocating limited attentions to limited coexisting online events, which has been supported by our big data from the Weibo platform (N=197,985). For the micro-level, we use agent-based modeling to verify the behavioral rules of Netizens. We set two agent categories (Netizens & Hots), which behave and interact autonomously. Based on multiple-round simulations, we obtained optimal solution outcomes. Besides, we have implemented mutual validation between big data and agent-based modeling, which has not been considered before. Three big data sets have been applied to check both validity and robustness of our solutions. For the set A, we use 89 online cases of the Shaanxi Province in China. For the set B, we used 147 online cases of the whole China in 2021. For the set C, we used 138 online cases under specific topic (the pandemic). It suggests that three optimal solutions have both validity and robustness. As three big data sets have different levels, cases, topics and durations, the model’s generality can be well supported. Hence, general behavior pattern of the Netizens can be revealed, which is of great significance for investigating human behavior online. Moreover, our work contains rich managerial implications, which helps to improve social simulation and prediction. Also, our TDS analysis framework can support or inspire other researches.