Since the re-outbreak of the covid-19 this year, a large number of emergencies and doubtful messages have been circulating on social platforms. This article focus on how risk society works during the epidemic on social media. To observe whether the government's communication strategies lower people's uncertainty toward the epidemic, we used the social media insight tool “CrowdTangle” to collect data from Facebook pages including “Ministry of Health and Welfare”, six mayors from special municipalities, and a well-known bulletin board “PTT Gossiping”. This paper process a self-built corpus, containing text data from fan pages mentioned above by using the Chinese independent corpus analysis tool “Corpro”. The result shows that although the government did not choose the unsuccessful deficit model to communicate with people, it still failed to lower people's uncertainty due to audience preference shift. The risk communication strategies become more complex in an environment where the audience can reach sufficient information online.
