Human Control of a Simulated Economic System

Abstract Fourteen teams were each asked to control a computer model of the British economy. For each time epoch of the task, they decided the rates of government expenditure, tax and limit on the money supply. Following the decision, they were provided with a set of economic indicators showing the consequences, and asked to decide for the next time period, and so on. Questionnaires on simplistic isolated economic relationships showed no benefit from the experience of controlling the model; nevertheless, the actual performance of each team ultimately corresponded to the predictions that the team had made in advance. Thus the importance of interaction with such a simulated system appears to lie in the ability to handle complex interactions and lags, rather than in verbally staled knowledge.