Detecting unintended consequences in engineering systems: What can we learn from multi-scale modeling?

Engineering systems such as transportation, healthcare and energy systems provide critical functions for modern society, but they are growing in both technical and social complexity. As a result, unintended design and policy consequences are becoming more common. Some of these consequences are the aftermath of systems being in the vicinity of critical thresholds, which can trigger abrupt transitions from one system state to another. In models, these thresholds correspond to bifurcations in their predicted behavior. In this paper we investigate whether multi-scale modeling, i.e. the use of distinct models at different scales, has the potential to aid in the discovery of bifurcations in engineering systems. To accomplish this, we reviewed and analyzed the literature on how multi-scale modeling is applied in the physical sciences including biology, material sciences, and physics. What we found is that despite their intended use for discovery, multi-scale models are largely constructed by tuning linkages among the scales to reproduce critical thresholds that have already been discovered empirically. This suggests the need for additional research to develop a systematic approach to introduce hypothesized linkage relationships to generate candidate bifurcations in our models. Such an approach could enable computational models to better inform engineers, managers and policy makers of the range of potential consequences of design and policy decisions.