By means of the semantical and proof-theoretical analysis of universal, high probability, and normic conditionals that we have given in part III, we have gained further insight into the (first- and second-order) notions of absolute and high reliability, and by that also into the justification of monotonic and nonmonotonic inferences. In this fourth part we are going to make use of the results of part III, together with the concepts and results of parts I and II, when we deal with the following questions: (i) is there a low-level agent which is ideal in the sense of section 8.6? (ii) If yes: what might the cognitive architecture of such an ideal agent look like? In particular: how may the typical properties of justified nonmonotonic inferences be implemented, i.e., the nonmonotonicity effect (that holds for nonmonotonic inferences in general) on the one hand, and the “optimum instability” , the closure properties, and the specifity sensitiveness (of justified nonmonotonic inferences) on the other hand?
KeywordsTuring Machine Symbolic Computation Belief State Cognitive Agent Soft Constraint
Unable to display preview. Download preview PDF.