Authors: JQ Smith, MJ Barons and M Leonelli
Title: Decision focused inference on networked probabilistic systems: with applications to food security
Abstract: Bayesian technologies have been progressively applied to larger and larger domains. Here, necessarily, probability judgments are made collaboratively and it is rare that one agent owns all the judgments in the system. So interesting new foundational and methodological issues have arisen associated with the status of inference supported by combinations of such judgments. In this paper we review some recent work on Bayesian inference underlying integrated decision support for huge processes. We argue that in a surprising number of such dynamic environments it is in fact formally justified to distribute the inference between different panels of experts and then use an agreed framework to knit these separate judgments to properly score different policies. We also briefly report recent progress in applying this new technology to develop an integrating decision support system
for local government officials to use when trying to evaluate implications on food poverty of shocks in the food supply chain if various ameliorating policies are applied.
KeyWords: Bayesian inference, Common knowledge, Decision support.