Prior studies on group recommendation systems usually explore heuristics to apply some interesting strategies in recommendations, for example, PloyLens [17] adopts least misery (i.e., the least satisfied user's preference) as the main strategy for its group recommendations. Aiming to better understand the latent factors behind group activities, in our study, we propose a probabilistic model to capture both of a user's own preference and her influence (termed as personal impact) to a group.