Natalie S. Glance, Tad Hogg and Bernardo A. Huberman
Dynamics of Computation Group
Xerox Palo Alto Research Center
Palo Alto, CA 94304
hogg@parc.xerox.com
a preliminary version appeared at AAAI Spring Symposium, 1994 a revised version to appear in Organization Science
We present a two-level model of organizational training and agent
production. Managers decide whether or not to train based on both the
costs of training compared to the benefits and on their expectations and
observations of the number of other firms that also train. Managers also
take into account the sum of their employees' contributions and the
average tenure length within their organization. Employees decide
whether or not to contribute to production based on their expectations
as to how other employees will act. Trained workers learn over time and
fold their increased productivity into their decision whether or not to
contribute. We find that the dynamical behavior at the two levels is
closely coupled: the evolution of the industry over time depends not
only on the characteristics of training programs, learning curves, and
cost-benefit analyses, but on the vagaries of chance as well. For
example, in one case, the double dilemma can be resolved for the
industry as a whole and productivity then increases steadily over time.
In another, the organizational level dilemma may remain unresolved and
workers may contribute at fluctuating levels. In this case the overall
productivity stays low. We also find a correlation between high
productivity and low turnover and show that a small increase in training
rates can lead to explosive growth in productivity.
postcript (323K)