Subjective opinions generalize probability distributions by including degrees of uncertainty which reflect lack of confidence in the probabilities. This paper describes a method for computing the joint subjective opinion of two variables which can be generalized to a method for computing joint subjective opinions over multiple variables in a subjective Bayesian network. We show how the joint opinions can be marginalized to provide subjective opinions on a reduced number of variables. With an example we compare the marginalization of a joint opinion with subjective logic deduction which also produces a marginal opinion.
展开▼
