Collaborative, Decentralized Engineering Design at the Edge of Rationality

Abstract

One perspective of a design process in the engineering design community is that it is largely a process marked and defined by a series of decisions. The fundamental assumption in most developed design decision support methodologies is that decision makers make rational choices; that is, choices that maximize the payoff for the predicted outcome. Decisions that do not maximize the predicted payoff are termed as mistakes or irrational choices and discarded. However, research in behavioral economics, psychology, and cognitive science has studied the human mind and suggested the notion of “bounded rationality” to explain decision errors. Bounded rationality refers to the intrinsic inability of human beings to accurately choose “rational” options prescribed by decision models such as expected utility. This paper extends the notion of bounded rationality within engineering design. Specifically, this paper studies the design of complex systems that require interaction among several different subsystems contributing to the overall product design. For convergent decentralized design problems, rational designers converge to equilibrium solutions that lie at the intersection of their individual rational reaction sets. These equilibrium solutions are usually not Pareto optimal and due to the dynamics of the designers’ interaction in collaborative design, it is rarely possible for them to converge to Pareto optimal solutions. However, when models for bounded rationality are introduced into individual designer behavior, it is seen that the converged solutions can improve the resulting solution. Bounded rational decisions within decentralized design are modeled, and the effects of propagating such decisions within a design process are studied.

Separating Design Optimization Problems for Bounded Rational Designers

“…This paper contributes to our knowledge about design decision-making by proposing specific random search algorithms as models for the choices of a bounded rational designer (unlike [40], who used a random error term). This idea agrees with (and comes from) the importance of search in design (cf.…”

“…Gurnani and Lewis [40] studied collaborative, decentralized design processes in which the models of the individual decision-makers (the designers) were based on the ideas of bounded rationality. In their model, the value chosen by each designer was determined by randomly sampling from a distribution around the (locally) "optimal" solution.…”

“…These algorithms should not be compared to state-of-the-art techniques for solving design optimization problems. Like [40], we are modeling the designer's bounded rational design choices as a random process. In [40], the values chosen by the designer were determined by a random error term.…”

“…Like [40], we are modeling the designer's bounded rational design choices as a random process. In [40], the values chosen by the designer were determined by a random error term. In our approach, on the other hand, the values are chosen by a random search process.…”

Towards a Theory for Unintended Consequences in Engineering Design

“…The authors propose that an unanticipated unintended consequence -what designers don't knowis fundamentally the result of the bounded rationality of designers. Bounded rational designers have limited time and mental resources and therefore may not have sufficient information to be able to make perfectly rational design decisions (Gurnani and Lewis, 2008). Simon proposed a number of reasons that bounded rational humans may not always make rational decisions (cited in Gurnani and Lewis, 2008).…”