Decentralized design at the edge of rationality

The design of complex engineering systems requires interaction among different disciplinary subsystems that affect the entire system. However, this kind of distribution or decentralization of the design decisions leads to designers having to make decisions under various sources of uncertainty and risk. Because of these uncertainties, errors could potentially creep into rational decisions made by designers. In this paper, we investigate the rationality of decisions made by engineers in a distributed design framework. For inherently convergent distributed problems, designers, when behaving rationally, converge to Nash Equilibrium solutions that lie at the intersection of their individual Rational Reaction Sets (RRS). However, these solutions are usually not Pareto optimal and due to the dynamics of the designers' interaction in distributed design, it is rarely possible for them to converge to a solution on the Pareto set. However, when error is introduced into individual designer behavior, it is seen that the converged solutions can be an improvement to the Nash solution. Further, inherently divergent distributed problems can even possibly converge when decision error is introduced. This paper studies the effects of the propagation of this error within a distributed design framework and demonstrates that errors made within "rational" decisions could possibly result in favorable solutions.