Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uunet!gistdev!bill
From: bill@gistdev.gist.com (Bill Busen)
Newsgroups: comp.ai
Subject: Re: scripts revisted...
Message-ID: <1160@gistdev.gist.com>
Date: 2 Apr 91 00:25:33 GMT
References: <9103290133.AA27890@enuxha.eas.asu.edu>
Organization: Global Information Systems Technology Inc., Savoy, IL
Lines: 17

gomez@ENUXHA.EAS.ASU.EDU (Jose L. Gomez-Rubio) writes:

>I think that the script representation can't handle unexpected scenarios and
>is only designed for stereotypical situations.

An approach that Gerald DeJong suggested to me to handle this is "script
switching and interaction" schema.  Let's say a fire breaks out in a
restaurant.  The restaurant script may say that you normally eat and pay before
leaving, and draw unkind inferences if you don't.  What you want is to
recognize that you are entering a multiple-script situation, determine that
the "fire" script bears more immediately on your goals (this approach requires
that you relate scripts to goals), and switch scripts before you burn your
posterior.  To me, this has the advantage of fairly well modelling what people
in fact do.

The strength of it is that the myriad unexpected situations that can occur
while tooling along in a script do fall into classifiable (scriptable) areas.