Abstract:  One of the major challenges in fostering a virtual community is the supply of knowledge, more specifically, the willingness and behavior to share knowledge with other members. Most virtual communities exhibited the participation inequality phenomenon that 90% of users were lurkers who never contributed, 9% contributed a little, and 1% contributed most of the messages (Nielsen, 2006). The bystander effect — the influence of the lurkers on other virtual community users’ knowledge sharing behavior, however, is very common but under-investigated. We proposed that the bystander effect would hinder knowledge sharing in virtual communities. Moreover, it was hypothesized that the bystander effect would decrease when the level of emergency and concreteness of posts thematic in virtual communities increased. Three situational experiments were conducted to examine the hypotheses. Experiment 1 studied the bystander effect on knowledge sharing in virtual communities using a one-factor between-subjects design. One-hundred twenty-one university students participated in the experiment. Experiment 2 examined the moderation effect of the emergency of posts thematic. A total of 177 university students participated in an experiment with a 2 (Number of bystanders: 1 vs. 54) × 2 (Emergency level of posts thematic: urgent vs. not urgent) between-subjects design. Experiment 3 examined the moderation effect of the concreteness of posts thematic. A total of 136 university students participated in an experiment with a 2 (Number of bystanders: 1 vs. 54) × 2 (Concreteness level of posts thematic: concrete vs. not concrete) between-subjects design. The results of these three experiments provided support for our hypothesis: (1) There was a bystander effect in the virtual community knowledge sharing: participants in the fewer-bystander (0 or 1) condition shared more knowledge in the virtual community than the more-bystander (14 or 54) condition. (2) Experiment 2 and Experiment 3 indicated that the bystander effect could be relieved or even reversed if the posts thematic appeared to be urgent or concrete. In the non-urgent or non-concrete condition, the classic bystander effect was found. Participants in the 1-bystander condition shared knowledge significantly more than participants in the 54-bystander condition. But when the posts thematic were made to be urgent, the bystander effect was reversed: Participants in the 1-bystander condition posted fewer messages than in the 54-bystander condition (Experiment 2). Similarly, when the posts thematic were made concrete, the bystander effect was relieved: In the 1-bystander condition, participants shared the same quantity of knowledge as in the 54-bystander condition (Experiment 3). The theoretical contributions and managerial implications of our findings were discussed. First, we broadened the understanding of bystander effect in virtual community knowledge sharing, which supports the social influence model and the non-linear relationship between the number of bystanders and knowledge sharing behaviors. Second, we demonstrated that changing the emergency and the concreteness of posts thematic could effectively change the bystander effect in virtual community knowledge sharing. This was an empirical evidence for the arousal: cost-reward model and the social influence theory. Third, results offered critical insights for managers. To change participation inequality and promote knowledge sharing, it is advised that managers should undo displaying the number of lurkers and ask users to concretely describe their questions with details in virtual community.

Key words:  virtual community, knowledge sharing, bystander effect, post thematic characteristics