Position Paper: AID'98 Strategic Knowledge in Design Computing Workshop


Routineness, Creativity and Strategic Knowledge
A Position Paper


In creative work it is not the prescriptive routine knowledge that differentiates between poor and good performance, it is the knowledge about what approach to take, what to do to recover from a problem, when to change tack or even when to take a break. Thus we need to understand how strategic knowledge can be represented, used and managed. [Edmonds 1998]

In [Brown 1996] I propose viewing ``routineness'' as a key dimension of design activity. Routineness depends on an individual's knowledge as well as their collected and processed experiences with types of design problems and classes of design requirements for each type. Routineness reflects the ability to match a situation to appropriate knowledge, as well as the appropriateness of the knowledge selected, and its ``ease of use''.

Another key dimension is the level of abstraction, which is tightly coupled with the type of decision being made. Hence, at the more abstract level (Conceptual design) the functional description is being refined, while at the more detailed end of the dimension (Parametric design) the parameters are being given values.

The Routine to Non-Routine and the Conceptual to Parametric dimensions provide a two dimensional space. Extreme points are RC, RP, NRC and NRP design situations.

Creative design activity [Goel 1997] is normally seen as starting at the NRC. That is, the lack of directly applicable knowledge and experience requires starting at a conceptual level.

Creativity is relative to a standard, to expectations. The performance standard can be that individual (or group) or an appropriate population. In addition, there is creative designing and there are creative designs (i.e., the activity or the result). Creativity can be discussed with respect to (wrt) both:



Individual Population
Activity A wrt I A wrt P
Result R wrt I R wrt P


A wrt I: in this category the strategy used by the designer needs to be new, or at least unusual in some way, relative to the designer's own, normal activity for problems of this kind. This results in creative activity.

A wrt P: in this category the individual's activity is judged relative to the population's normal approaches; i.e., standard, recommended methods. The individual's actual activity need not be creative by the A wrt I standards, even though one would probably expect it to be so. The result need not be new with respect to the population standards either -- a known design could be produced by a new method.

R wrt I: in this category the resulting design is new or unusual for the designer. It is possible that routine knowledge could produce this. This might occur if a previously unused combination of techniques were needed.

R wrt P: in this category the resulting design is new or unusual with respect to the population's designs. While it's also possible that routine knowledge could produce this, as routineness is relative to an individual's experiences, it is much less likely.


As routine design knowledge is not merely algorithmic, strategic and often heuristic choices still need to be made. Punch et al [1995] describe the knowledge-based selection mechanism that was developed for DSPL [Brown & Chandrasekaran 1989].

That mechanism depends on situation matchers (Sponsors) that evaluate and propose the approach that they `represent'. Those evaluations are reviewed by a Selector that uses additional knowledge, including the design history, to pick an approach.

This Sponsor/Selector mechanism has been used in systems to switch between both small-scale and routine activities (e.g., design plans), and also between large-scale reasoning tasks (e.g., classification).

Creative designs or creative design activity can be fostered by a willingness to try different approaches, a deliberate seeking for associations between the current situation, and deliberate priming with related (but not necessarily relevant) knowledge.

This should lead the designer to select alternative and perhaps less favored approaches -- i.e., the Selector knowledge is deliberately adjusted. It should also allow the use of less routine approaches via analogical matching, instead of basic CBR or plan selection.

It doesn't mean that routine knowledge will not be used, but rather that it won't be used all the time, and that it may be used it slightly different situations than normal.


References

D. C. Brown, Routineness Revisited. Mechanical Design: Theory and Methodology, (Eds.) M. Waldron & K. Waldron, Springer-Verlag, 1996, pp. 195-208.

D. C. Brown & B. Chandrasekaran, Design Problem Solving: Knowledge Structures and Control Strategies, Research Notes in Artificial Intelligence Series, Pitman Publishing, Ltd., London, England, 1989.

E. Edmonds, Call for Papers, Strategic Knowledge in Design Computing Workshop, AID'98, http://bashful.lboro.ac.uk/aid-wshop/, 1998.

A. K. Goel, Design, Analogy, and Creativity, IEEE Expert, Vol. 12, No. 3, May/June 1997.

W. F. Punch, A. K. Goel & D. C. Brown, A Knowledge-Based Selection Mechanism for Strategic Control with Application in Design, Assembly, and Planning. International Journal of Artificial Intelligence Tools, Vol. 4 (3), Sept. 1995, pp. 323-348.


Biographical Information


Version: Wed May 6 21:22:40 EDT 1998