Weighted Logic

While logic provides an appealing formalism for knowledge representation and reasoning, it is not without drawbacks. There are many aspects of human reasoning that are not adequately handled by pure logic. (Of course there are also many aspects that we would rather avoid capturing.) For example, suppose a human is told that “all aliens are blue”. Later, however, the human also learns the fact that “aliens of type Klingon are not blue”. While this new fact is contradictory to the original statement, a human would have no difficulty resolving the conflict. Most humans would treat aliens of type Klingon as a special case, and otherwise all other aliens will be believed to be blue. Now consider how an automated reasoning system would handle this same situation. Suppose that the KB already has knowledge that no alien can be both blue and green, i.e.