Large-scale models have very important theoretical, technical, ethical, and political costs and consequences that are often not recognized, and when recognized, often ignored. These costs and consequences prevent large-scale models from playing a constructive role in policy processes and retard the development of a more useful largescale modeling activity. This essay discusses some of these costs and consequences and recommends their explicit recognition and acknowledgments as the first steps for improvement.KEY WORDS: society, decision making, large-scale models, social systems. w HE CREATION and use of large-scale T models to understand and master society's many problems have not been entirely successful. There are many complicated reasons for this lack of success, too many to cover in a short essay. Emphasized here are the costs and consequences of failures in large-scale modeling more than the causes of failure; the central question here is, "What are some longer-term costs and more fundamental consequences of relying on large-scale modeling?" Addressing this question is more modest, but potentially more important, than simply drawing up another list of the travails and shortcomings of "big models" projects. Many such lists have been produced. A representative sample, and no more than that, would include: Ascher (1978);Szanton (1981);Greenberger, Crenson, and Crissey (1976);Brewer (1973);Brewer and Shubik (1979);Greenberger, Brewer, Hogan, and Russell (1983); Pugh (1977); Deutsch et al. (1977).The direct costs of modeling in terms of man-hours, computer time, data expenses, and so on, may be significant, but will not be considered. Nor are we too concerned about the mountains of computer excreta such efforts produce, especially in light of the mole hills of genuine insight the mountains represent. Ascher's book (1978) contains perhaps the most comprehensive (and devastating) assessment of a diverse range of modeling and forecasting experiences in the literature. A striking and consistent
I66Behaviors1 Science, Vulume 2R. 1983 finding in his study is that when genuine insight has occurred it is has most often been in spite of the mountains of heavy analytic work, not because of them. His work is both important and disturbing.This essay is primarily concerned with other not so well understood effects related to theoretical, technical, ethical, and political costs and consequences-many of which extend well beyond the immediate settings of the large-scale social models themselves. These four conceptual categories were first proposed in Brewer (1973) in appraising urban models and then subsequently employed in Brewer and Shubik (1979) for models in the national security/military area.Before sketching some of these costs and consequences, we will review briefly the usual justifications for building large-scale models, the main field and areas of application, and the common experiences.