Probabilistic approaches to non-monotonic reasoning

It has been found that the ability to handle incomplete information or to perform nonmonotonic reasoning does not exist in some probabilistic reasoning mechanisms such as INFERNO and probabilistic logic. Two relevant issues in this area are examined: why these mechanisms cannot be used to perform nonmonotonic reasoning and how this capability can be implemented in them. It is shown that difficulties with these mechanisms are due to their monotonic nature. The author proposes methods for implementing this type of reasoning in these mechanisms. Some implications of this work and possible future research are discussed.<
>