Evolutionary Game Models on Multiagent Collaborative Mechanism in Responsible Innovation

Abstract

Innovation is a game process; in particular, the behavior among multiple agents in responsible innovation is susceptible to the influence of benefits, risks, responsibilities, and other factors, resulting in unstable collaborative relationships. Therefore, this paper constructs a tripartite evolutionary game model including the government, enterprises, and the public, combined with system dynamics modeling to simulate and analyze the tripartite behavior strategy and sensitivity to relevant exogenous variables. The study shows that the tripartite game eventually converges to a stable state of the government active supervision, enterprises making responsible innovation, and the public’s positive participation. The positive participation of the public drives rapidly the game to a steady state, while the behavioral strategies of enterprises are more susceptible to the behavior of the government. Supervision cost, penalty amount, and value compensation are the most critical factors influencing the change of the corresponding agents’ behavior strategy, and the final strategic stability of tripartite is affected by multiple exogenous variables.

Tripartite Evolutionary Game Analysis of Participants’ Behaviors in Technological Innovation of Mega Construction Projects under Risk Orientation

“…The evolutionary game theory takes the group as the research object and reveals the complexity and dynamic evolution path of the behavior from the perspective of finite rationality, which is widely used in the analysis of network cooperation, enterprise cooperative innovation, the stability of strategic alliances, etc. [32][33][34].…”

Experience Weighted Learning in Multiagent Systems

“…MARL has been proved to be advantageous in at least three aspects: parallel computation, distributed layout, and communication between agents. Although agents in MAS have higher capability to adapt to a complex environment, several challenges arise in the real world [20]. For example, QMIX [21] and VDN [22] that incorporate deep reinforcement learning are designed for team tasks with networks.…”