What makes a simulation useful? [TRANSIMS]

Modern computers make possible a new blending of systems, man, and cybernetics in the detailed simulation of large sociotechnical systems. Several such simulations are currently under development at Los Alamos National Laboratory and elsewhere. When deployed, they will affect the daily lives of hundreds of millions of people and the allocation of billions of dollars. Whether they are deployed depends entirely on their perceived usefulness, which in turn depends on answers to the following: What kinds of questions does the simulation address and what kinds of solutions does it provide? How can the solutions be validated? Is simulation more cost-effective than other methods? Answers to these questions lead us to define a useful simulation as one which efficiently provides correct, robust estimates required by decision-making needs, together with well understood variability for the outcomes in hypothetical situations. The paper examines the implications of this criterion for the design of TRANSIMS, a regional transportation network simulation system, and by extension, for simulations of other sociotechnical systems.