Complex socio-ecological systems/Other Annotated Bibliography entries

March, James G. 1991. Exploration and Exploitation in Organizational Learning. Organization Science 2(1): 71-87

This paper explores a formal model of group learning. It models knowledge as a series of correct or incorrect beliefs towards m dimensions of reality. It models an "organizational norm" set of beliefs, and a series of individuals who each have their own set of beliefs. In the model, individual belief sets converge towards the organizational norm, and the organizational norm is influenced by the "more correct" group of individuals. This seems like a very artificial situation, but it is representative of the methodology of formal models and simulation. Despite its artificiality, exploration of the model reveals some interesting characteristics. In all simulations, the individual knowledge sets and the organizational norms converge over time. However, if learning by individuals is slow, the model yields a higher equilibrium level of knowledge than if individual learning is fast. This is because there is more time for the "more correct" individuals to influence the organizational norm, before all individuals converge to that norm. A related result is that having a mix of slow and fast learners yields a better final equilibrium level of knowledge than a uniform, intermediate rate of learning. And, having turnover in the group of individuals (e.g. as people leave an organization and are replaced by newcomers) gives a better final result, especially if the rate at which newcomers converge towards the organizational norm is slow.

March then extrapolates from the model of organizational learning to the broader issue of "exploration" and "exploitation." Exploitation would be converging on current best practice; specialization on what you know to be the best current solution. Exploration means investing resources to look for other possible solutions; the cost of this is that you know that over the short-term you are moving away from what you know works best now. In terms of a fitness landscape, exploitation is moving towards or staying at the local optimum; exploration is moving away from the local optimum to find a higher fitness peak elsewhere in the landscape. Organizations (and all agents in systems) must make tradeoffs between "investments in learning" (exploration) and "consumption of the fruits of current capabilities" (exploitation).

Schelling, Thomas C. 1978. Micromotives and macrobehavior. Norton and Company. 252 pp.

Schelling is a winner of the Nobel Prize in Economics for his game-theory analyses of the cold war. In this volume, he uses simple models to help understand "social dilemmas" -- the often unexpected group outcomes that result from the sum of individual choice. A significant part of his research looked at racial sorting between neighborhoods. He shows that a small preference by individuals to avoid being in an extreme minority can result in a process that inexorably leads to complete racial separation, even though that result is far from the actual preference of any individuals. The result comes from individual decisions to optimize with very minor constraints, which replicate throughout the system to produce an "emergent property" of the system that is far removed from any individual's desired outcomes. There are many examples, some of them apparently trivial: if no one wants to be in the least skillful decile of a tennis club, then those who are in this decile will leave; as they do, the next decile becomes the lowest, and they too leave, until there is no one left in the club.