Sequential Decision-Making Under Uncertainty Using Hybrid Probability-Possibility Functions

Didier Dubois , Hélène Fargier , Romain Guillaume , Agnès Rico

Probabilistic and possibilistic models of sequential decision problems are known to possess good behavioral and algorithmic properties. In this paper, the range of models of problems of sequential decision under uncertainty that are dynamically consistent, consequentialist and allow for tree reduction is enlarged by considering a representation of uncertainty that is both probabilistic and possibilistic. The corresponding utility functional is expected utility for highly likely states, and an optimistic or pessimistic possibility-based criterion for unlikely states.

Mots clés: Decision tree; Possibility theory; Fuzzy measures; Belief functions

Domaines: Informatique [cs]; Intelligence artificielle [cs.AI]

Didier Dubois, Hélène Fargier, Romain Guillaume, Agnès Rico. Sequential Decision-Making Under Uncertainty Using Hybrid Probability-Possibility Functions. 18th International Conference on Modeling Decisions for Artificial Intelligence (MDAI 2021), Sep 2021, Umeå (virtual), Sweden. pp.54-66, ⟨10.1007/978-3-030-85529-1_5⟩. ⟨hal-03382413⟩