Non-monotonicity in probabilistic reasoning

We start by defining an approach to non-monotonic prob­ abilistic reasoning in terms of non-monotonic categorical reasoning. We identify a type of non-monotonic proba­ bilistic reasoning, akin to default inheritance, that seems to be commonly found in practice. We formulate this in terms of the Maximization of Conditional Independence (MCI), and identify a variety of applications for this sort of default. We propose a formalization using Point­ wise Circumscription. We compare MCI to Maximum Entropy, another kind of non-monotonic principle , and conclude by raising a. number of open questions.