Abstract: Sending comments on bullet screen video websites is one form of herd information comment behavior. This paper uses the method of semi-structured interview to collect data from users of bullet screen video websites. Grounded theory and NVivo11 are employed to conduct three levels of data coding. During the open coding, typical statements in the interviews are selected. Free nodes are formed by analyzing and generalizing tacit concepts. Child nodes are then formed by the categorization of the free nodes. During the axial coding, tree nodes are generalized by defining and analyzing the child nodes. During the selective coding, herd information comment behavior is conceptualized as a second-order construct based on the interview data. The relationships between herd information comment behavior and other tree nodes are also proposed. Consequently, a theoretical model regarding driving factors of herd information comment behavior by the users of bullet screen video websites is established.

Key words: Herd information comment behavior；Bullet screen video websites；Grounded theory；Driving factors, User behavior analysis