A study of the effects of agent activeness on team performance

Passive agents participate in team activities passively, that is, only upon request, whereas active agents involve themselves voluntarily. Teams composed of active agents are generally believed to perform better than those with passive agents. In this paper, by using a computational simulation model we examine the effect of agent activeness on the efficiency of decision-making teams that access different amount of information. ‘Team-Soar’ is a computational framework that consists of a group of interconnected individual AI agents (i.e., Soar). A simulation experiment using Team-Soar was performed. Results of the simulation provide valuable insights on the roles of agent activeness. For example, the impact of having more active agents becomes more significant as the amount of information to process increases and when the team decision efficiency is important. Some of the results are counter-intuitive and therefore provide an opportunity to understand the roles of the agent activeness more deeply. For instance, the simulation results reveal that having more active agents did not always enhance team efficiency. Conclusively, the simulation experiment demonstrates how computational models contribute to the research of agents' social characteristics.