When a Model Needs a Model: Convincing Skeptics of Counterintuitive Results—An Example from Affirmative Action Modeling

Abstract

Counterintuitive results constitute one of the important benefits of computer simulation models. Yet such results are often viewed skeptically because by definition they confound our expectations, and our expectations are one measure of the validity of the model. The task of the model builder, then, is to convince skeptics of the validity of unexpected outcomes and thereby reinforce the credibility of the model. One way to do so is to appeal to an abstraction that is even simpler and better understood than the abstraction that the simulation model represents. In the case discussed here—the issue of how worker layoffs are determined-the counterintuitive results of the model are explained through a "thought" model involving urns and marbles, a scheme familiar to most people who have taken a course in elementary statistics (and easily understood by those who have not suffered through such a course). Keywords: simulation, models, validity, counterintuitive results, social policy, affirmative action, personnel.