Position Paper: AID'98 Emergence in Design Workshop
A Position Paper
David C. Brown
AI in Design Research Group,
Computer Science Dept.,
WPI, Worcester, MA 01609, USA.
An identifiable design property which has not been explicitly represented can be said to be emergent. Discovering and using the emergent properties in design is regarded as emergence in design. [Chase & Schmidt 1998]
First, let me suggest a rewording of the above implied definition that I think is an important refinement:
An identifiable design property which has not been explicitly anticipated or explicitly represented in the current (partial) design can be said to be emergent.
If a property was ``anticipated'' it wouldn't have that `surprise' quality that's characteristic of emergence. That is, if you knew it was coming, it isn't a surprise, even if it isn't explicitly represented.
However, I realize that this a weak distinction. In the classic example of four squares being placed together to produce an extra square that they enclose, that extra square still appears, even if you know it's coming. Perhaps we might call that Expected Emergence?
If a property has ``not been explicitly represented'' anywhere in the available knowledge of the person or system then it cannot be directly ``identifiable''. Putting three squares together to enclose a triangle, for example, isn't significant if you aren't able to recognize it as a triangle. That is, you need to already have knowledge of triangles. If you can match the emergent design property directly with existing knowledge, let's call this Directly Identifiable emergence.
It is possible to imagine a discovery process that identifies a particular property of a design as ``interesting'' -- to use the AM term [Lenat 1982] -- and therefore worth remembering, classifying and naming. For example, with no prior knowledge of triangles you might be able to discover that emergent shape and label it. Whether you could use it once found depends on whether what you discover is classifiable, and whether the properties of it's type can be inherited. Let's call this Indirectly Identifiable emergence.
Clearly, identification can be done deliberately, perhaps driven by a goal that encourages active openess during the design process -- such as might occur when someone is trying hard to be creative, as opposed to just `going through the motions'. In this situation one can imagine both directly and indirectly identifiable emergence: people would take the time to use discovery processes.
Identification can also be done in a more passive/unconcious way, where associations between stored knowledge and the ongoing design trigger matching, leading to directly identifiable emergence. Such `noticing' requires some `priming', either by explicitly making a mental note [Adelson & Soloway 1985] or by having recently used knowledge available due to some other task or subtask.
The matching present in identification is also part of the Analogical Reasoning process, an important aspect of creativity in design [Goel 1997]. However, it's argued that for analogy the matching requires abstraction in order to be possible. For example, to make an analogy between pipes and wires you need to have appropriate abstractions of both water and electricity so that they match sufficiently.
It's hard to imagine this being done as automatically as a ``see a square, know it's a square'' match. Hence, analogical reasoning might be one of the discovery processes that provide indirectly identifiable emergence.
In addition to analogical reasoning, functional reasoning [Umeda & Tomiyama 1997] (i.e., more conceptual level reasoning) is seen as a key component of creativity -- although you can be ``routine'' at any level [Brown 1996].
Functional reasoning in design is concerned with the intended use of the design, its purpose. While the analogical use is interesting (e.g., a shoe used as a hammer, or a pen used to make a hole in a piece of paper) it isn't designed in. It's usually discovered, due to a need, by analogical reasoning (and possibly by experimentation). One might consider this to be a kind of emergent functionality, as it is revealed after the design has been instantiated.
In [Balazs & Brown 1998] we use analogy to reason about function. For our work we consider function to be the use of a design's properties by the environment.
The ``environment'' might be a person, or some other system. The word ``use'' is intended to imply that there is some purpose for the interaction, and ``property'' is defined as loosely as it has been above. For example, a property might be:
- a `state change', such as a movement,
- `shape', such a sharp edge,
- `rigidity, or
Under what circumstances would a function be emergent during the design process? Probably ...
An identifiable function of a design which has not been explicitly anticipated or explicitly represented in the current (partial) design can be said to be emergent.
Of the examples above, it's not hard to imagine indirectly identifiable emergence of function due to shape, color, rigidity or weight. And `shape' is clearly capable of providing directly identifiable emergence of function, in an analogous fashion to the four squares example given above.
However, the extra issue here is to establish ``use'', or more precisely the capability of use. One obvious possibility is that these are noticed during mental simulations of the design that are intended to check whether the desired properties and behaviors of the design have been successfully incorporated.
Any additional functionality noticed can lead to the retrieval and use of new knowledge, in a fashion similar to analogy. For example, if one notices that the designed object when used is likely to leave some sort of mark, and that's not a bad thing, then one might consider incorporating some of the structural or behavioral characteristics of pens.
Consequently, having an emergent function is not only possible, but may even be beneficial, especially if creativity is being sought.
B. Adelson & E. Soloway, The Role of Domain Experience in Software Design, Trans. on Software Engineering, IEEE, Vol. SE-11, No. 11, 1985, pp. 1351-1360.
M. E. Balazs & D. C. Brown, A Preliminary Investigation of Design Simplification by Analogy, Proc. Artificial Intelligence in Design '98, Lisbon, Portugal, July 1998.
D. C. Brown, Routineness Revisited. Mechanical Design: Theory and Methodology, (Eds.) M. Waldron & K. Waldron, Springer-Verlag, 1996, pp. 195-208.
S. Chase & L. Schmidt, Call for Papers, Emergence in Design workshop, AID98, http://www.arch.usyd.edu.au/~scott/AID98/emergence-workshop/ Lisbon, 1998.
A. Goel, Design, Analogy, and Creativity, IEEE Expert, Vol. 12, No. 3, May/June 1997.
D. B. Lenat, AM: Discovery in Mathematics as Heuristic Search. Knowledge-Based Systems in Artificial Intelligence, (Eds) R. Davis & D. B. Lenat, McGraw-Hill, 1982, pp. 3-225.
Y. Umeda & T. Tomiyama, Functional Reasoning in Design, IEEE Expert, Vol. 12, No. 2, March/April 1997.
Version: Wed May 13 21:37:52 EDT 1998