for the scissors/rocks/paper game
although you might be able to predict the "en masse" average
behavior better in the unconstrained case you would be less likely to
predict the specific behavior. In fact, you probably would only
correctly predict the specific behavior 1/3 of the time.
In contrast, in the constrained case, fairly simply models of the
actor/agent/person should predict specific behavior more than 1/3 of the time.
Further, it is likely that you might be able to even predict en masse behavioer
better.
In a related vein -
we had data on interact at two points in time from a tailor shop in Zambia.
Here, the individuals are constrained - by their knowledge, education,
religion, tribal norms, ethnic heritage, position in the tailor shop etc.
A random model of interaction did fairly well, and predicted a large
number of the cases where people continued to interact/stopped interacting/
started to interact/ or never intereacted. However, a relatively
simply model that took into account the knowledge network in which
individuals were embedded = produced better predictions.
Kathleen
p.s. embedded -
I use the term to indicate that individuals do not exist in a vacuum
but rather are connected to others (agents, robots, webbots, groups,
institutions, organizations, etc.) through webs of interaction and
knowledge, assignment to tasks, shared goals, etc. These webs or networks
are seen as mutually constraining and enabling.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|