The principal-agent theory focuses on designing incentive mechanisms to solve the problem of moral hazard under the condition of asymmetric information. However, there are some limitations in the current studies of principal-agent problem. On the one hand, although much effort has been made to apply principal-agent theory to specific relation structure, most of previous studies mainly focus on single principal-agent relations. On the consideration of some participants have double identity of principal and agent in reality, the multiple principal-agent relations is more truthful than classic principal-agent theory. On the other hand, the previous theories are developed under the assumption of relational economic man. With the developing of behavioral economics, a growing number of scholars believe that people's behavior is bounded rational rather than purely rational. The theory of process fairness preference considered that process fairness preference is an important factor affecting human behavior, which provides a theory basis for the view of bounded rational.Beware of two disadvantages of the traditional principal-agent theory, this paper make a rewarding approach to classic theory through constructing chain-like multi principal-agent relations with incorporates fairness preference. Which has extended the traditional principal-agent theory research.In the third part of this paper, taking the double principal-agent for example, the agent in the first stage of principal-agent model is the principal of the second stage principal-agent model, thus there is a modification on principal-agent structure. Through considering the double principal-agent that pure principal, intermediary and pure agent are entirely rational, it has found that, under the condition of entirely rational, the optimal level of effort of intermediary and pure agent is positively related to the optimal revenue sharing ratio. In each stage principal-agent model, when the principal gives the agent more incentive the agent will try harder to work, but the degree of risk aversion of the agent will make the incentive effect weak. Therefore managers of enterprises could take good use of the method of increasing revenue sharing ratio to fully arouse the work enthusiasm of the employees. However, the greater the degree of risk aversion of the agent, the result of through increasing revenue sharing ratio coefficient to motivate agent is less obvious, so the employees who have a smaller degree of risk aversion are good employees.In the fourth part, this paper constructs a chain-like double principal-agent model through inducing process fairness preference. After calculation, we've got more profitable conclusion that the agent is friendly to the principal if the agent's effort is greater than the half of his biggest effort level, and the principal is friendly to the agent if the fixed income which the principal gives the agent is greater than half of the maximum. At the same time, the more fixed income, the more effort agent will give. On the one hand, it has well explained the Rabin's thought of “give a plum in return for a peach” and “an eye for an eye”, on the other hand, it also shows us the reason that employees have better performance and professional ethic in some companies which have good welfare policy.In the fifth part, through numerical simulation, this paper has found that it has an equilibrium which make every participant has more revenue. In conclusion, optimal incentive contract under the entirely rational can be improved, and process fairness preference can increase the income of all the players.The analysis in this paper can also be applied to design an incentive mechanism for enterprises, financial institution, and government budget problem as well as profit distribution for up and down stream of supply chain. Therefor it would have significant practice in real economic.