Dealing with Subjective Uncertainty in Knowledge Based Systems

Knowledge based systems depend on algorithms able to relate the inputs of a system to a correct answer coming out of the knowledge-base. Practical systems shows that imperfect information will always get into the data-base and an imperfect knowledge-base will always exist, thus it is usual for a knowledge based system has to be able to model and deal with information imperfections. The kind of information imperfections involved in know ledge-base systems can be modeled by the "subjective uncertainty ", which with the "objective uncertainty " compose the dual nature of the uncertainty, a taxonomy first defined by Helton. One of the formal models that deals with subjective uncertainty is the Mathematical Theory of Evidence, or Dempster-Shafer Theory. This theory provides a method for combining evidence from different sources without prior knowledge of their distributions, however, it has some pitfalls caused by the non natural embodiment of the uncertainty in the results. In this paper we analyze the counter-intuitive behavior of the theory, identify the sources of the subjective uncertainty, and present a method of automatic embodiment of the uncertainty which overcomes the pitfalls, allowing its use in a broad range of situations.