EAGER: Hybrid Socio-Technical Teams: A Theoretical Framework for Modeling and Design of Hybrid Networks of Human and Autonomous Agents

Stevens Institute Of Technology, Hoboken NJ

Investigators

Abstract

Many critical systems in the future will rely on coordinated teamwork of a hybrid group of humans and smart, decision-making, autonomous agents. Such hybrid teams are becoming increasingly common in a wide variety of socio-technical systems. Members of such teams are highly heterogeneous, work in rapidly changing and uncertain environments, have different goals and motivations, and are often spatially distributed with little direct face-to-face interaction. This creates serious challenges for team-level coordination and cooperation, and can result in decreased efficiency, or even disastrous failure, in some critical cases. This award supports fundamental research to provide needed understanding of dynamics, behavior, and coordination mechanisms of human-agent networks. Outcomes of this project will result in better understanding of the behavior of such networks, and more efficient design of future intelligent systems for disaster response, energy and transportation, all of which will in the future rely on efficient, dynamic coordination between a group of people and a network of autonomous agents. Understanding hybrid teams and addressing their coordination challenges require an interdisciplinary approach that combines computational and behavioral sciences. To this end, this research will combine multi-agent computational methods such as complex network analysis and game theory with social psychology of team behavior. The research team will create parametric, context-dependent, analytical and computational models to understand how team-level coordination and cooperative behavior are affected by variables such as composition of team members, communication network structure among nodes, distribution of humans and machines on networks nodes, and key parameters in the environment. Analytical methods, such as game theory on graphs, and verifiable agent-based simulations will be used to draw causal inferences from the network structure and agent compositions to overall team-level performance parameters. The research will also test the hypothesis that the connectivity structure of the team network can be used as an effective control parameter to dynamically steer the cooperative behavior of hybrid socio-technical systems. To validate the developed models, the research team will perform a data-driven study of disaster response teams. It will investigate the impact on overall coordination and efficiency when introducing autonomous agents into human disaster-response teams.