next up previous index
Next: Knowledge packets Up: Communication Previous: Knowledge structures

Explicit messages

  In addition to observing another supertask's memory structures, supertasks communicate by means of explicit messages passed between them. In the ISAAC model, there are four types of control         messages--suggestions, anticipations, expectations, and requests. During normal processing, there may be times when specific information is required which a control message can acquire. Some situations will require that normal processing be interrupted; the necessary information is important enough to warrant this. Other situations will allow normal processing to occur, perhaps with some modification. Crossed with these two possibilities is what should happen if the information being requested is not discovered. If the information is not discovered, the reasoner may elect to handle this in a special fashion or to ignore the missing information. These two axes produce four possibilities:

These three messages are sent from a supertask when it discovers information. There is one additional type of message that is sent by a supertask when it needs to explicitly ask another supertask for information (as such, it does not fit into the two-by-two grid):

 

        Requests, then, are questions which a supertask needs answers to; suggestions, anticipations, and expectations are some of the ways in which these answers may be provided.

    The idea of expectation-driven processing is certainly nothing new; a great deal of AI research, in fact, makes use of some variation of it (e.g., see [#!ai:handbook1!#,#!expect:birnbaum-selfridge-1981!#,#!read:ram1!#,#!expect:riesbeck-1975!#,#!read:schank4!#]). However, the         result of this wide usage has been that the exact meaning of the term has become clouded. My use of the terms ``expectation'', ``suggestion,'' and ``anticipation'' attempts to clarify some of the nuances which have become attached to the overused term ``expectation'' through its years of use.  


next up previous index
Next: Knowledge packets Up: Communication Previous: Knowledge structures
Kenneth Moorman
11/4/1997