There is often a trade-off between system efficiency (macro-level) and network stability (micro-level) when designing multi-agent networks. As more system components incorporate autonomous/AI agents whose decisions interact with those of human agents, it becomes essential to analyze the nature of the trade-off between efficiency and stability — and explore conditions needed to simultaneously achieve both by selecting the right network architecture. The goal of this presentation is to utilize strategic network formation models to determine the efficiency and stability of network systems as a function of their structure and to model the nature and intensity of the trade-off between the two. Using heterogeneous connection cost parameters, we demonstrate how efficient networks typically follow a core-periphery architecture. Our first step is to characterize the exact solution for a number of important classes of networks. Finally, we discuss the stability conditions and show that in general, there are a large number of stable structures and that, under the strict notion of pairwise stability, the efficient network is unstable. Our model then incorporates direct utility transfer between pairs of agents and shows that core-periphery networks have the unique property of simultaneously achieving scalability and stability for a crucial class of network structures. Moving on to multi-core and peripheral systems, we discuss efficiency and stability in such systems and how they can be achieved. We will conclude the presentation by discussing different engineering and social implications of the work and offering empirical examples to support the findings.

BIO

Babak Heydari is an associate professor at Northeastern University’s Department of Mechanical and Industrial Engineering, as well as an affiliate faculty member at the School of Public Policy, the Network Science Institute, and the Institute for Experiential AI. In his interdisciplinary research, he uses network science to study the architecture of socio-technical and human-AI systems and its relationship to collective behavior and social norms, platform-based sharing economy systems, and co-evolution of structure and behavior in complex systems. After receiving his Masters and PhD from the Department of Electrical Engineering and Computer Science at UC Berkeley, he spent three years working with start-ups in Silicon Valley. He has been the PI and Co-PI of several projects sponsored through NSF, DARPA, INCOSE, SERC and a number of private corporations. Prof. Heydari is a CAREER Award recipient from the National Science Foundation.