Learning from data by coherent probabilistic reasoning

We apply the theory of (de Finetti) coherent inference to the handling of uncertainty in the process of automatic medical diagnosis. Given some possible diseases (which could explain an initial piece of information) and a tentative probability assessment of them, the database consists of conditional probabilities P(E|K), where each K is a disease and each evidence E comes from a suitable test. The coherence of the whole assessment is checked. The doctor can now update the probability of each disease and check again coherence of the whole assessment, since the diseases do not constitute, in general, a partition (so that the usual Bayes theorem cannot be applied). These steps can be iterated until a degree of belief sufficient to make a diagnosis is reached: the coherence condition acts as a control tool on every stage