Does strategy need computer experimentation?

The method of computer experimentation, which explores complex social phenomena using artificial adaptive agents, has attracted a growing interest in the field of strategy. Because the strategy problem domain is characterized by high levels of difficulty, uncertainty and complexity, traditional methodological approaches may not always provide the most efficient protocol for discovering successful theories. New computational methods, in which tractability constraints are less binding than in their purely mathematical counterparts, appear to be promising tools for the strategy researcher. We examine how and why these tools are likely to advance the strategy agenda. Contrary to the criticisms of some, we argue that these methods do, indeed, constitute a valid form of social science. However, computer experimentation is no silver bullet: unique limitations accompany its potential benefits. We discuss these and conclude that the new computational methods are most fruitfully viewed as complements to, not substitutes for, their more traditional scientific counterparts.