Keywords
scenarios; optimization; decision making under deep uncertainty; exploratory modelling
Start Date
7-7-2022 2:00 PM
End Date
7-7-2022 2:20 PM
Abstract
Complex systems such as cities, energy grids, or the global climate have many possible futures. Narratives of decision-relevant futures (scenarios) are a common tool for making the complexity and uncertainties of complex systems humanly interpretable. They can present contrasting plausible futures, summarize alternative pathways, and serve as foils for evaluating policy options. However, the effectiveness of the policy evaluation depends in part on selecting a set of scenarios that captures the full diversity of the system’s plausible future developments. Here we show an optimization-based approach for generating scenarios that are specifically designed to be diverse, plausible, and comprehensive. We establish the advantages of our method by evaluating on Schelling’s segregation model against three previously proposed methods for developing scenarios: generic archetypes, scenario matrices, and clustering. Importantly, we also find that our method can reveal previously unrecognized futures, which may bring emergent events or “black swans” to the attention of decision-makers. Finally, we observe that existing methods may have substantial shortcomings when applied to nonlinear systems, as they fail to cover the full spectrum of plausible futures. Our results show that computational scenario generation can support decision making under uncertainty by explicitly searching for insightful scenarios using optimization techniques. The resulting scenarios can then be used to provide a more insightful and robust basis for policy decisions, especially for complex systems with emergent behavior, or where substantial uncertainties are present.
Finding Diverse Future Scenarios for Complex Systems
Complex systems such as cities, energy grids, or the global climate have many possible futures. Narratives of decision-relevant futures (scenarios) are a common tool for making the complexity and uncertainties of complex systems humanly interpretable. They can present contrasting plausible futures, summarize alternative pathways, and serve as foils for evaluating policy options. However, the effectiveness of the policy evaluation depends in part on selecting a set of scenarios that captures the full diversity of the system’s plausible future developments. Here we show an optimization-based approach for generating scenarios that are specifically designed to be diverse, plausible, and comprehensive. We establish the advantages of our method by evaluating on Schelling’s segregation model against three previously proposed methods for developing scenarios: generic archetypes, scenario matrices, and clustering. Importantly, we also find that our method can reveal previously unrecognized futures, which may bring emergent events or “black swans” to the attention of decision-makers. Finally, we observe that existing methods may have substantial shortcomings when applied to nonlinear systems, as they fail to cover the full spectrum of plausible futures. Our results show that computational scenario generation can support decision making under uncertainty by explicitly searching for insightful scenarios using optimization techniques. The resulting scenarios can then be used to provide a more insightful and robust basis for policy decisions, especially for complex systems with emergent behavior, or where substantial uncertainties are present.
Stream and Session
false