A methodology is proposed by which the design of a knowledge based system (KBS) facilitates the evolution into a more structured script-based approach. Scripts consist of semi-structured (in the sense of Gorry and Scott Morton, 1971) knowledge about stereotypical events. Scripts comprise a set of roles, props, goals, locations, and events. The process and methodologies used to move from an ill-structured problem to a running KBS and subsequent evolution toward a script-based situation are sketched. The methodological approach is grounded in an illustration with an expert system designed to assist entrepreneurs with lease negotiations. From their experience with LEASE, a GURU-implemented expert system (ES), the authors derive several heuristics to assist in the process of transforming an ES into a script. They also comment upon specific features of the leasing task, LEASE, and the resulting script: a wide variety of variables affect the process of enscripting. Sample heuristics include: examine the KBSs knowledge tree to look for patterns; select KBSs to enscript that run on problems that recur frequently and problems that are salient and important to the user; avoid KBSs that involved a great deal of subjectivity.<<ETX>>
[1]
Virginia E. Barker,et al.
Expert systems for configuration at Digital: XCON and beyond
,
1989,
Commun. ACM.
[2]
J. Casey,et al.
Picking the right expert system application
,
1989
.
[3]
John P. McDermott,et al.
R1: A Rule-Based Configurer of Computer Systems
,
1982,
Artif. Intell..
[4]
John J. Sviokla,et al.
Putting expert systems to work
,
1988
.
[5]
Pamela K. Coats.
Why Expert Systems Fail
,
1988
.
[6]
W. R. King,et al.
Achieving the potential of decision support systems.
,
1983,
The Journal of business strategy.