What drew me to look at parti diagrams were the many sessions during my own architectural education where a favorite critic of mine insisted that I draw a diagram when beginning to explain my ideas at a desk crit rather than using the actually drawings of the proposal. During those episodes, I drew many of these diagrams trying to describe what I really wanted to show with those heavily worked on sketches piled high on my desk. This tactic not only helped me to explain the concepts of my design, but it also helped my critic in trying to extract my dominant concerns and to give alternative solutions and ways of seeing it while adhering to my overall concept. He also showed me ways of looking at the diagram that would provoke architectural solutions that would not surface if we had only reflected on the drawing and not the diagram. He often found inconsistencies between the drawing, the diagram, and the words that I used. In locating these inconsistencies, I could resolve them or use them as a design idea. These experiences showed me the utility of the diagram as an inference and checking mechanism. The abstraction and ambiguity in these simple concise diagrams demonstrated their strength by the multiplicity of design instances that could be generated from it. To me, it became analogous to a poem of a few verses. Those few verses could give very vivid images and meanings that may be interpreted differently at different times. With added verses, one can get a more detailed sense.
2. I have spent much time thinking and talking about some abstract notion of chunking architectural information in order to use it as a checking mechanism for generations of designs. In the need to extract architectural knowledge, I became engrossed in trying to think about what is architectural knowledge. This side-tracked me. I think, however, that it gave me a start in looking at what might be the type of operators through which I can funnel the parti diagram for the extraction of relevant knowledge. (This also opened up the mouth of the big whale and in I went directly in the belly!)
3. About developing models: The best (or most optimal) computational models for a task may not emulate human performance at all. I think of the past desire to make a computational chess opponent. The best computer chess programs actually go through a brute force search of the possible moves and perform an optimization on it. In building such a system, we may not necessarily learn anything that is actually happening in the mind of the human chess-player. I think that there is something more complex than a brute force search. This to say that chunking may not be computationally optimal, but it does have some parallels to how we may perceive/remember/process certain types of information.
I have, however, began to formalize in my own mind how I may use parti diagrams and how multiple representations of architectural knowledge differing in degrees of abstraction and ambiguity can be mechanisms for creativity.