Theory of Endorsements and Reasoning with Uncertainty.

摘要

The general knowledge acquisition problem and the problem of acquiring strategic knowledge from experts was addressed. An automated knowledge acquisition tool called ASK was described and demonstrated with a human machine dialog, and the results from experiments analyzed. The importance of the design of knowledge representations and reasoning methods was emphasized since it plays a central role in the knowledge acquisition process. Reasoning under uncertainty has two aspects. One is to assess the most likely states of the world, the other is to act on those assessments. The former is often called judgement and the latter decision-making. Judgement has been the primary focus of research on reasoning under uncertainty in AI, while decision-making (lately these have been called planning problems) which deals with how autonomous agents act in uncertain environments, is increasingly gaining more attention. An adaptive planner called PLASTYC was built to operate in a dynamic, spatially-distributed, multi-agent, ongoing, unpredictable, and real-time world simulator for controlling forest fires. The simulator, called Phoenix, was used as a framework to discover functional relationships between environment characteristics, autonomous agents' behaviors, and agents' designs. (kr)