Scenario-based reasoning and probabilistic models for decision support

This paper presents a scenario-based approach to deal with uncertainties in situation assessment problems. Scenario representation is based on causal models, whereas scenario generation involves the estimation of the states of model variables, done by means of observations and inferences of hidden states by using domain knowledge. Moreover, scenario management is addressed by means of a probabilistic framework involving Bayesian and credal networks, which allows the evaluation and ranking of scenarios according to likelihood, used to prioritize information to be presented to decision makers. The presented scenario approach also supports the adaptation of the reasoning models on the fly, as scenarios are generated and relevant information changes or becomes available.