rolandi@gollum.UUCP (Walter Rolandi) (09/10/88)
In response to the learned O'Keefe .... >So, when designing a learning system, how do we deal with this? >How should it "decide" when to look for an explanation of something, >and when to change the subject? A learning system simulating animal behavior might look for an explanation only when something is missing. "Something missing" can be represented with scripts which have slots containing those events that are typically present across instances of environmental events. These slots can be populated by means of a simple feature abstraction algorithm. Such a system might support normal inferencing and possibly question generation. Now changing the subject is another matter all together. It may be a function of the number of "explanation searches" generated that lead to either nothing or some unhappy consequence. Walter Rolandi rolandi@ncrcae.UUCP rolandi@ncrcae.Columbia.NCR.COM NCR Advanced Systems Development, Columbia, SC