Abstract:
Decision making (DM) is a targeted choice of actions based on given knowledge and preferences. Normatively, Bayesian DM, maximising expected utility, should be used under uncertainty but this happens less than desirable. Often, imperfection of the DM participant can be blamed as it limits the deliberation effort spent. In economical, societal, biological and technical systems,
labour division copes with complex DM tasks within which each participant selfishly solves DM tasks it manages, while interacting in market fashion with its similar neighbours. Project will equip them with logically complete and conceptually feasible DM support, which will increase mildly their deliberation load while requiring no perfect facilitator or unlimited external
resources. The support will employ fully probabilistic design (FPD, a generalisation of Bayesian DM) for: elicitation of domain
knowledge and preferences into probabilities used by FPD; merging of incomplete and incompletely compatible distributions, offered by neighbours; feasible approximate learning and design of DM strategies.