Communities of practice are difficult to understand. When trying to build a model to represent or simulate some domain, it is necessary to understand both the theory of the domain, and the internal knowledge of its practitioners. The duality of this echoes back to the difference between information and knowledge, theory and practice. There are two types of knowing, knowing-that and knowing-how. When building models to be run by a computer, the first of these is rather easy to encode, but the second is much more problematic. The issue of practice opens a bag of worms that has no explicit or clear solution.

It is this issue of learning the practice of a domain that I am interested in. Much of the theory that comes about attempting to learn and document these domains involves developing layers of abstraction and performing a psychological study (usually through interviews and observation) to learn how the practitioners actually practice. I think this approach is useful, but flawed. I think that a lot may be gained through more anthropological means. In order to understand and model, the analyst must become a practitioner in the domain itself.

One example that I think is relevant in that case is the idea of the artist-programmer, the need for which has been emphasized by Janet Murray, Michael Mateas, Ken Perlin, and Mary Flanagan, among others. This is especially relevant in the games industry, where students coming from one discipline or another are usually taught to interact with the other half, but rarely are artists taught programming or programmers taught art. It is my opinion that this has led to a somewhat conceptual divide, preventing real communication and effective interaction.

Additionally, many communities of practice are often influenced only in limited ways by theory. They might develop a consistent framework of practice that may be interpreted as theory, but the practice may not have been informed by theory in the first place. Even in disciplines where all there is is theory, issues of practice become very important. An example of this is in the InTEL project, where we have developed a model of Statics in general, but more specifically, we have developed a model of one means of practicing it, originating from the instructors who we have been working with. This suggests that, in practice, there is no theory. Theory is what one might call a derivative model.

This idea actually makes a lot of sense when compared against a lot of critiques of classical AI. Dreyfus in particular emphasizes that human reasoning is all pattern recognition, and in essence, is all practice. He argues that theory enters the picture when a novice first begins to learn a skill, and is no longer referenced afterwards. I do not agree completely with his argument, but the idea of focusing on practice has merit. This may be used as a critique of models that examine theory in whole, but that does not mean that models are worthless. One may still build models of practice.

This introduces another dimension, though. If one cannot model theory in general, then one must model practice. However, practice is an individual phenomenon. No two practitioners in a domain will have the same approach to ideas and problem solving. As a result, modeling is a fundamentally creative act.