How to draw plausible conclusions from uncertain and conflicting sources
of evidence is one of the major intellectual challenges of Artificial
Intelligence. It is a prerequisite of the smart technology needed to
help humans cope with the information explosion of the modern world. In
addition, computational modelling of uncertain reasoning is a key to
understanding human rationality. Previous computational accounts of
uncertain reasoning have fallen into two camps: purely symbolic and
numeric. This book represents a major advance by presenting a unifying
framework which unites these opposing camps. The Incidence Calculus can
be viewed as both a symbolic and a numeric mechanism. Numeric values are
assigned indirectly to evidence via the possible worlds in which that
evidence is true. This facilitates purely symbolic reasoning using the
possible worlds and numeric reasoning via the probabilities of those
possible worlds. Moreover, the indirect assignment solves some difficult
technical problems, like the combinat ion of dependent sources of
evidcence, which had defeated earlier mechanisms. Weiru Liu generalises
the Incidence Calculus and then compares it to a succes sion of earlier
computational mechanisms for uncertain reasoning: Dempster-Shafer
Theory, Assumption-Based Truth Maintenance, Probabilis- tic Logic, Rough
Sets, etc. She shows how each of them is represented and interpreted in
Incidence Calculus. The consequence is a unified mechanism which
includes both symbolic and numeric mechanisms as special cases. It
provides a bridge between symbolic and numeric approaches, retaining the
advantages of both and overcoming some of their disadvantages.
