First, let me thank everyone who has contributed to this thread. Your replies mean a lot to me and confirm my impression that social network analysts are among the most collegial people on earth.
It will take some time to read the recommended materials and to think through their implications for what I am trying to do. At this point, however, none seem to address the problems I have in mind when I talk about network chemistry.
Consider a joke circulating on the Internet.
"Two chemists walk into a bar.
The first says, I'll have a glass of H20.
He is handed a glass of water and drinks it.
The second says, I'll have a glass of H20, too.
He is handed a glass of hydrogen peroxide (H202).
He drinks it and falls down dead."
Similar scenarios are common when project teams are formed to do creative work. Some combinations result in work that is adequate but bland. Others result in work that is poisonous. Only rarely is the work exciting enough to be celebrated at advertising contests like the one from whose annuals I take my data. How do network properties affect these results?
Thinking about teams, I notice several features not found in the usual examples of network analysis and network simulation.
I find myself trying to imagine how to construct an agent-based model to simulate what my historical and ethnographic data tell me about the Japanese ad world.
I start with the individual actors who make up teams.I assume the relevance of safety, effectance, and status seeking as motives. All of these actors will score relatively high on effectance. They are, after all, advertising creatives. We will, however, see variation in likelihood to play it safe. Status seekers will score especially high on effectance. But there is more to be considered. We also need a list of skills, approximated, given the data we have, by the roles they are able to play in a team. And, then, there is the question of talent, a random variable.
Plus, we will need to consider the process by which actors are assembled into teams. Availability, access, required skill(s): what else do we need? Does previous success working together strengthen the likelihood of working together again? How much is it offset by shifting demand for specific skills, itself the result of changing market conditions? Our ethnographic data include both individuals who have worked together for years and appear to be close friends and other individuals who say that ability is everything, nothing else is relevant, "We don't have to live with these people."
How far can we go before the mathematical and computational issues become intractably difficult? That is not an area in which I have the relevant expertise.
It does seem clear to me that currently available network flow and network architecture models do not address these issues. I would like to be proved wrong.
These, anyway, are my current rough thoughts. Additional feedback helping me to refine them will be much appreciated.