Joe Roussos (London School of Economics), Roman Frigg (London School of Economics), Richard Bradley (London School of Economics)
Increasingly many policy decisions take input from collections of scientific models. Such decisions face significant, poorly understood uncertainty. We rework the recently developed "confidence" decision theory to tackle decision-making with model ensembles, showing how it can be used to construct nested sets of predictions of increasing specificity. We discuss the conditions under which particular sets are available to decision-makers. We illustrate the approach with a case study: an insurance pricing decision using hurricane models. The confidence approach has important consequences for hurricane insurance, and we generalise these to a wide variety of policymaking contexts.
