In the AI in Design community, it is a commonly held view that we now know how to write knowledge-based systems for routine design problems. While that probably depends on the domain, and on the exact notion of routineness, that sounds right to me.

Almost all of the community would also agree that the ability to build design systems that display Creativity [Boden 1994b] is not yet within our grasp. However, as we feel that designing is implementable using a computer, and therefore that a sizeable chunk of human reasoning is expressible in this fashion, I suspect that the majority would side with Boden's ``scientifically-minded individuals'':

Many scientifically-minded individuals would argue that there is nothing especially problematic about creativity. A creative idea, they would say, is merely a novel (and valuable) combination of familiar ideas. Accordingly, creativity could be explained by a scientific theory showing how such novel combinations can come about [Boden 1994a].

How do we get a knowledge-based design system to ``reason outside the box''? What exactly do we mean by ``the box'', or ``outside'' for that matter?

Despite a limited amount of existing work (for example, see [Gero & Maher 1993]) the field of AI in Design doesn't have a scientific theory. With the notable exception of McLaughlin's paper [1993], most of the papers in that book offer neither a precise statement about what creativity is, nor pointers to the psychological literature.

If we don't have a clear, detailed definition of what properties we are trying to produce, then how can we tell if our systems display them?

There's a huge, and quite popular literature about enhancing and ``releasing'' one's creativity (see the links at [Cave 1999] for example). A good introduction to the more serious, psychological literature is Boden's book [1994b], and Eysenck's paper in particular [1994].

Eysenck points to ``private novelty'' -- the discoveries of an individual that are new to that individual -- and ``public novelty'' -- new to everyone. Boden, in a similar but not identical definition, calls these P-creativity and H-creativity, for Psychological and Historical respectively.

He does not appear to make the distinction between creative designing (process) and creative designs (product), i.e., the designer's process itself may be creative [Brown 98].

Eysenck points to studies that suggest that creative individuals have a cognitive style that is called ``overinclusive''. This leads to unusual word or problem associations, unusual sortings of items, and some ``loosening of associative thinking''.

This suggests that creative individuals utilize such abilities in order to reason ``outside the box'', but use their strong domain knowledge and intelligence to critically assess the novel concepts that are generated. This cognitive style encourages more associations, and would lead to natural use of analogy.

There has been some work on the use of analogy to support knowledge-based creative design systems [Goel 1997]. However, there's need to explore some of the more subtle aspects of reasoning, such as overinclusive reasoning. There's no need to suppose that any one large reasoning mechanism will provide us with creative systems. We need to look deeper.

References

Boden, M. A., (1994a). Agents & Creativity, Communications of the Association for Computing Machinery, special issue on Agents (ed. D. Riecken), available as http://www.ozemail.com.au/~caveman/Creative/Authors/mbagents.htm

Boden, M. A., Editor (1994b). Dimensions of Creativity, MIT Press.

Brown, D. C. (1998). Routineness, Creativity and Strategic Knowledge, Proc. AID'98 Strategic Knowledge in Design Computing Workshop, http://www.cs.wpi.edu/~dcb/Papers/AID98-strategic.html

Cave, C. (1999) Creativity Web. http://www.ozemail.com.au/~caveman/Creative/index.html

Eysenck, H. J. (1994). The Measurement of Creativity. In: Boden, 1994, pp. 199-242.

Gero, J. S. & Maher, M. L., Editors (1993). Modeling Creativity and Knowledge-based Design, LEA.

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