Stephen R. Haptonstahl

For the foreseeable future, my research will have two foci: a substantive question, and the development of methods to help answer this substantive question.

Can the Central Intelligence Agency reorganize to avoid a 9/11 style failure? More generally, can a bureaucratic information-gathering hierarchy (BIGH) be designed so as to make the probability of catastrophic error arbitrarily small? Based on my years of experience in this kind of organization -- tactical crew aboard a deployed AEGIS cruiser -- I conjecture is that this minimal-error BIGH is not possible.

A purely informational approach, akin to the Condorcet Jury Theorem, suggests that a minimal-error BIGH is achievable simply by having a sufficiently flat and wide hierarchy. However, a purely informational characterization of the problem ignores two key characteristics of BIGHs. First, BIGHs are composed of people with cognitive limitations, including limits as to the number of subordinates one can manage effectively and as to the amount of information one can consider simultaneously. Both of these limits place bounds on the width of a hierarchy. Second, these people have their own individual goals, such as career advancement, job security, expansion of resources, prestige, and compliance with professional norms. These goals can conflict with the goals of the organization, which means there is a need for superiors to create incentives for subordinates to support the goals of the organization.

Modeling cognitive limitations means stepping back from the assumption of fully rational actors. Modeling incentives in a BIGH means implementing some form of principal-agent model specifically to characterize information aggregation.

To address these modeling needs, I am working on several sets of tools. Modeling cognitive limitations effectively means using both analytic and computational formal modeling techniques. I am working with other computational modelers inside and outside political science to develop and adopt rigorous standards for making scientific arguments using computational models. To build these models to capture the important characteristics of bureaucrat behavior, I am testing specific theoretical assumptions using lab experiments. To facilitate testing of models, I am extending the quantal response equilibrium (QRE) literature into models with continuous action spaces, especially principal-agent models.

Experiments in virtual worlds may prove to be vital for testing theories of institutional design. Massively multiplayer online roleplaying games offer the potential of large-scale experiments with a high degree of control. However, research into the economies of these virtual worlds -- and their exchange rates with the real world -- imply that participants derive real value from the results of their participation. This means that many or all of the same ethical considerations that apply to real-world field experiments must apply to these virtual field experiments. Regardless, virtual worlds offer greater institutional diversity than the real world, and thus provide a rich environment for exploring what otherwise might be counterfactual conditions.

The tools I develop are intended for a specific purpose, but will be widely applicable in and out of political science, and should advance, if marginally, the state of the art.