Opinion Dynamics In Online Review System

Due to the rapid growth of Internet,consumers obtain information about products and experiences from online rating platforms,which further influences their purchase decisions.Online reviews have a growing influence on the purchasing decisions of potential consumers.Therefore,the research on online reviews has attracted wide attention of scholars and has become an important research direction.From the perspective of opinion dynamics,more then three million reviews on Dianping.com were collected to analyze the internal mechanism in the online review process.The main objectives and results obtained of the thesis are listed as follows:(1)Considering the differences in the opinion interaction pattern,traditional opinion dynamics models cannot explain individual review behaviors.In order to explore the evolution process of opinions on the online review platform,we optimized the network structure and confidence thresholds of traditional models based on the collected data from Dianping.com,and created an opinion dynamics model for online rating platforms.The proposed method can advance the understanding for the evolution process of online opinion.(2)Influence of self-boosting manipulation on rating dynamics in online rating systems.This study employed a sequential approach of rating series that is context-free and computationally efficient to identify vendors with suspicious manipulated ratings,and identify the manipulation intervals.The results show a statistically nonsignificant difference in the average rating scores between before and after manipulation,indicating that manipulation does not influence follow-up rating dynamics.Moreover,the difference in average ratings between before and after manipulation decreases within a long time window,which indicates that the long-term effect is weaker than the short-term one.To cross-validate the result of the effects of rating manipulation,we further conducted a field experiment on Dianping.com to manually provide fake ratings and examined the effect of the manipulated rating on the follow-up rating dynamics.To further reveal the underlying mechanisms of why the follow-up raters can correct the bias of manipulation,an agent-based model was constructed by simulating users’rating decision-making process given the presence of high-value previous ratings.