Resolving the dilemma of responsibility in multi-agent flow networks

Deshmukh, Jayati, Adivi, Nikitha and Srinivasa, Srinath (2023) Resolving the dilemma of responsibility in multi-agent flow networks. Mathieu, Philippe, Dignum, Frank, Novais, Paulo and De la Prieta, Fernando (eds.) In Advances in Practical Applications of Agents, Multi-Agent Systems, and Cognitive Mimetics. The PAAMS Collection. vol. 13955 LNAI, Springer Cham. pp. 76-87 . (doi:10.1007/978-3-031-37616-0_7).

Abstract

Multi-agent networks often face the “dilemma of responsibility” where optimising for individual utility may result in sub-optimal network-level outcomes. But, imposing constraints on individual agents for obtaining better network-level indicators, may severely impede their utilities and rationale for participating in the network. We address this problem of the conflict between individual utility and collective outcomes, using a decentralised approach called Computational Transcendence (CT) which is based on modelling agents with an elastic sense of self. We discuss how this model can be applied to realistic multi-agent application scenarios. The first scenario is on decision-making in multi-agent supply chains, and the second is on adaptive signalling in a road network. In both these applications, we compare CT with several baseline models and find improvements across multiple application-specific metrics. CT is shown to outperform strategies for individual utility maximisation, by improving network-level indicators in an emergent manner, without posing a high burden of responsibility on individual agents.

This record has no associated files available for download.

Contributors

Author: Jayati Deshmukh

Download statistics