Judgmental forecasting and anticipation in human process control

The task of the human operator in industrial process control is usually to keep a system within specific limits or to optimize some parameters. These activities are thought to imply a strong anticipation component: decisions take into account some assessment of the future course of events. But the nature of that assessment is currently not well defined; an explicit and statistically optimal prediction is only one possibility. In agreement with the literature on judgmental forecasting, we show in three experiments that, when asked to predict future values of time series, naive subjects seem to use a representativeness heuristic that is not statistically appropriate but demonstrates a real knowledge of some statistical characteristics of the series. We discuss a number of theoretical and practical consequences