Each student began a research project during the two-week workshop. Below are brief descriptions and links to these various projects. The expectation is that these projects will form the basis for dissertation chapters and/or journal articles.

Given the ambitious nature of the projects, this page is likely to change over the next few months as projects reach completion over the summer.

Susan is analyzing the adaptive dynamics invoked by various selection mechanisms on a simple model of competitive information sharing. She has found that simple changes in selection algorithms can alter the phase-space of the dynamics in such a way that very different compositions of final strategies emerge in the system.

John is formally modeled the strategic dynamics of coordination games like those discussed in Schelling and Arthur. He has found that the dynamics tend to promote ``inferior'' equilibria in these games.

Barry is investigating coalition formation in cooperative games. He is finding that the introduction of boundedly rational agents into such an environment results in new notions of coalition formation that, in many cases, allow very reasonable solutions to form. Such solutions can also be closely linked to some important existing notions of rational coalition formation.

Nick is investigating the simple adaptive dynamics of coordination games like those discussed in Schelling and Arthur. He will also be conducting laboratory experiments on these games.

Michael is using a genetic algorithm to find useful predictive patterns within voting data. Such nonlinear techniques may serve as a superior alternative to many existing statistical techniques.

Robert is exploring the dynamics of locally-connected agents playing the repeated Prisoner's dilemma. By adjusting both neighborhood size of interactions and adaptation, he is finding a variety of interesting results concerning the strategic dynamics of the system.

Abstract: Profit maximization is difficult. Sophisticated and experienced managers often disagree about which action is most likely to maximize profits for a given firm. Economic models of profit maximization, on the other hand, are--in general--easy. Well-trained economists can readily discern the action which maximizes the firm's objective function. The global maximum is unique and achievable because the objective function is designed to have this property. This paper weakens the assumption of analytically tractable objective functions. I propose a model of profit maximization in which it is, essentially, impossible for the firm to discover the global maximum. Firms have no choice but to, in the words of Lindblom (1959), "muddle through" in their attempt to find the optimal budgetary allocation in an extremely complex economic landscape. Computer simulations provide details of that landscape as well as evidence that certain strategies may be more effective in difficult environments like health care delivery. A copy of this paper can be obtained here.

Steve has developed a model in which agents desire a status good in
which their utility is derived by their consumption of a necessary good
(e.g., food) and their * relative rank * in consumption of the
status good with respect to their neighbors (e.g., vacations). Early
investigations indicate that the system can easily cycle with a series
of arms races forming and then slowly dissolving.

Kieron has begun to explore a model of competing organizations. Organizations can alter their structure so as to become more accurate information processors at the cost of increased processing time. The implications of such alterations are, of course, a function of the behavior of the organizational structure of competitors.

Elizabeth is exploring the dynamics of simple general equilibrium models using ``tunably-rational'' agents. While such agents do converge on Pareto optimal allocations, the resulting final allocations are far from the competitive equilibrium. She is current investigating the underlying factors responsible for this behavior.

Jordan is analyzing adaptive models of voting behavior when special interests can influence parties and preferences.

Michael is studying the generic behavior of boundedly rational agents attempting to optimize payoff on complex problem spaces. He has shown how search in these worlds can be modeled by a Markov process. His current work focuses on finding the full implications of such an analysis.

Joel has developed a model of trade across a geographic space in which endogenous trade networks can spontaneously form and dissolve.

John H. Miller , miller@zia.hss.cmu.edu.