Aggressive behavior has been a hot topic in the field of developmental psychology. One of the important problems in the field of aggressive behavior is that researchers neglect the influence of emotion on children's cognitive procession. In addition, there were few studies that probed the characteristics of decision-making of children who involved in aggression/victim. If researchers take the integrated process of emotion and cognition into considration when exploring children's decision-making, the results would reflect more truely and comprehensively the involved process of aggression/victim in the situation of social information processing. The gambling task paradigm of affective decision-making is thought to induce performance similar to the individual decision-making behavior in real life. A self-designed computer program of standard gambling task, in which 80 test items were divided into 4 blocks, was used therefore to compare decision-making achievement of children belonging to different aggression/victim groups in different periods from the process perspective so as to reveal the features of affective decision-making of children belonging to different aggression/victim groups. The above is the innovation of this study. Gambling task consisted of 80 test items, each involving four doors A, B, C and D, from left to right. The net losses of A and B are equal in general, so are C and D. But A or C belongs to the high-frequency small punishment, while B or D belongs to the low-frequency big punishment. In the long run, it is disadvantageous to choose A and B, but advantageous to choose C and D. Using the way of companion nomination and scales compiled by Masten and Salmivalli, the author first selected out 262 subjects, which were categorized into aggressors/victims and non-involved matching by grade and gender. The results show that there were obvious differences among different groups of children, comprising net difference of advantage-disadvantage selection, decision-making difference(style) of advantage-disadvantage selection and frequence of rewards and punishment. The performance of children in each block reveals the following conclusions. First, in net difference of advantage-disadvantage selection, the overall behavior of non-involved children is better than other groups of children with learning effect of performance gradually improved in the experimental process of decision making. Aggressive victim’s affective decision-making scores are worse than non-involved children’s, with improving prophase performance and unchanging anaphase performance. Indirect-aggressive children’s show strong learning effect during latter stage. Victim’s and direct-aggressive children’s decision-making scores are the worst in the overall trials, the former displaying the course of upgrading in prophase then descending in anaphase, and the latter exhibiting a little ascending of learning effect. Secondly, in acquirement style of advantage-disadvantage selection, all kinds of children manifest finally the trend of selecting more from advantage doors, but non-involved children the earliest display this tendency with the largest scope. Aggressive victim rank the second. Indirect-aggressive children show the last but display large scope afterwards. Victim’s and direct-aggressive children display comparatively smaller scope. Last, in frequency style （high-frequency small punishment vs. low-frequency of big punishment）, direct-aggressive children, aggressive victim children and non-involved children select more from choices of high-frequency but smaller punishment, whose risk is low and steady. The three kinds of children present punishment-oriented decision-making style. So, the gambling task paradigm of affective decision-making can distinguish different types of aggression/victim children, and children’s involving in aggression/victim behavior is closely related to their affective decision-making level and style.