Paper detail

Emoticon-based Ambivalent Expression: A Hidden Indicator for Unusual Behaviors in Weibo

Recent decades have witnessed online social media being a big-data window for quantificationally testifying conventional social theories and exploring much detailed human behavioral patterns. In this paper, by tracing the emoticon use in Weibo, a group of hidden "ambivalent users" are disclosed for frequently posting ambivalent tweets containing both positive and negative emotions. Further investigation reveals that this ambivalent expression could be a novel indicator of many unusual social behaviors. For instance, ambivalent users with the female as the majority like to make a sound in midnights or at weekends. They mention their close friends frequently in ambivalent tweets, which attract more replies and thus serve as a more private communication way. Ambivalent users also respond differently to public affairs from others and demonstrate more interests in entertainment and sports events. Moreover, the sentiment shift of words adopted in ambivalent tweets is more evident than usual and exhibits a clear "negative to positive" pattern. The above observations, though being promiscuous seemingly, actually point to the self regulation of negative mood in Weibo, which could find its base from the emotion management theories in sociology but makes an interesting extension to the online environment. Finally, as an interesting corollary, ambivalent users are found connected with compulsive buyers and turn out to be perfect targets for online marketing.