Quasi-Bayesian behaviour: a more realistic approach to decision making?

In this paper the theoretical and practical implications of dropping -from the basic Bayesian coherence principles- the assumption of comparability of every pair of acts is examined. The resulting theory is shown to be still perfectly coherent and has Bayesian theory as a particular case. In particular we question the need of weakening or ruling out some of the axioms that constitute the coherence principles; what are their practical implications; how this drive to the notion of partial information or partial uncertainty in a certain sense; how this partial information is combined with sample information and how this relates to Bayesian methods. We also point out the relation of this approach to rational behaviour with the more (and apparently unrelated) general notion of domination structures as applied to multicriteria decision making