Abstract

The antonym of (conditional) relevance is (conditional) irrelevance or independence. The chapter introduces this notion and develops its theory. It thus turns out that the full theory of Bayesian nets can be carried over to ranking theory. This results in a most illuminating (and extensively exemplified) algorithmic theory of the propagation of evidential impact within ranking nets and thus of inductive reasoning. An appendix shows how this theory extends to an account of (independent) testimonies (that is much simpler than its probabilistic counterpart).