Coherent conditional probability as a general inferential tool in AI

In this paper we deal with an approach to probabilistic logic based on the concept of coherence (which goes back to Bruno de Finetti). Many peculiarities of the concept of coherent conditional probability (which is based on a different approach with respect to the usual one in probability) entail a large flexibility in the management of any kind of uncertainty. In particular, in our framework it is possible to put together different kinds of partial knowledge, as in the paradigmatic example discussed in the second part of the paper