Online rating can directly reflect users’ collective behavioral patterns which is of great concern in online social systems. In this paper, we investigate the correlations between the users’ rating behaviors and the real-time updated average ratings of objects given from other users’ previous ratings. We average all the ratings rated after the real-time displayed average ratings at a given interval after dividing the data into five groups according to the user degrees. By analyzing two real systems, the results show that in general there is a linear correlation with slope one between them when the displayed average ratings are between 2.0 and 4.5, but users rate higher scores if the displayed average ratings are lower than 2.0, and give lower ratings if the average ratings are higher than 4.5. Besides, small-degree users would rate higher than the real-time displayed average ratings, while large-degree users are stricter with their ratings than the others so that they usually give lower ratings whatever the movies are. Furthermore, the distributions of the users’ rating bias in all the five groups show that the rating biases of the large-degree users are small, yet those of the small-degree users are relatively large. Our findings could be helpful to analyze online users’ collective behaviors as well as abnormal behaviors in the networks.
