Multi-Actor Value Modeling for Federated Systems

Abstract

Design methods for federated systems must consider local incentives and interactive effects among independent decision-makers. This paper extends value-centric design methodology to multi-actor cases by using game-theoretic principles. Federated systems are represented as a Stag Hunt game where players choose between independent (noncooperative) and federated (cooperative) strategies. Risk arises from the possibility of withdrawn cooperation and can be quantitatively assessed for particular design alternatives with subjective estimates of the probability for cooperation and objective measures such as the weighted average log measure of risk. The value of cooperation can be bounded below by designing for the independent strategy and above by designing for a centralized strategy controlled by a single authority. An application case considers a stylized system value model of a distributed satellite system with two players and symmetric design decisions. Results demonstrate lower and upper bounds to the value of a federated strategy with opportunistic, fixed-cost communication services. Quantitative risk analysis shows the value-maximizing federated design under mutual cooperation carries risk which can be mitigated by more conservative design alternatives with only minor loss of value.