Human society is a typical complex system. The emergence and evolution of many complex phenomena in human society are closely related to human decision-making behavior. Studying human decision-making behavior is helpful to understand the mechanism and principle behind the complex phenomena of human society, and it is of great guiding significance to improve the efficiency of social management. Although individuals in complex systems may seem to have various characteristics, they have internal connections. And some seemingly random changes follow simple rules. This paper mainly focuses on the group behavior under the complex system, through the Exploitation-Contribution game experiment, Attacker-Defender Contest game experiment and the corresponding theoretical modelings, to explore the individual behavior principle steam within the group and the causes of group behavior. 

    The Exploitation-Contribution game is the analogy to the behavior of exploitation and contribution in society and nature. Through behavior experiments under different group sizes, we found that when the relationship between exploiters and contributors is stable, contributors tend to cooperate and exploiters tend to be fair. Win-win cooperation between the two sides holds the dominant position. When the group size increases, the punishment of the TFT strategy is apportioned by the group size, resulting in a decrease in the penalty. Thus, as the size of the group increases, contributors' punitive measures have no obvious effect on exploiters, the proportion of distribution chosen by exploiters is getting lower and lower, and ultimately contributors have to cooperate and endure exploitation. Then, with the help of evolutionary game theory, we construct the payoff matrix of the basic exploitation and contribution game and join the TFT strategy and reputation to expand. Based on the expanded payoff matrix, we use the Moran process to perform numerical simulations using the Moran process and the replication dynamic equations that add "mutation rates". The numerical simulation results are a good validation of the behavioral experiment conclusions. 

    For the Attacker-Defender Contest game experiment, we explore the role of oxytocin in group conflict, and how team members rely on historical information to coordinate collective action in group conflict. We found that oxytocin enhances the level of cooperation of individuals and significantly improves the coordination of attack group members. Besides, the order of decision-making can affect individual behavioral decisions. In the simultaneous decision-protocol, the rear-positioned member determines whether to attack based on the contribution value of the leading member. At the same time, the contribution ranking of the intra-group members is very stable, attackers sit primarily on the contribution value of the group member in the last round, while defenders focus on the contribution value of the previous round of the opposing group and the contribution value of the member who contributed the most in the group in the previous round. In the experiments, we observed that the defenders' contribution is generally higher than the attackers'. To explain the behavioral differences brought about by the role differences, we use the best response theory to construct the theoretical model. The best response model is based on the experimental design, assuming that the people will make decisions according to the information of the previous round. We find that for defenders, the benefits of winning games are much higher than losing games, so they contribute more positively. For attackers, the winning or losing of the group conflicts with personal interests. Sometimes losing the game will get higher personal interests, so the attacker needs more trade-offs between win-loss and personal interests.