Using judgment to deal with streams of data: Forecasting and other tasks
People often have to use their judgment to deal with streams of data. In control tasks, they have to act in some way to ensure that the data stream has desired characteristics. For example, doctors treat people with drugs in an attempt to ensure that diagnostic indicators are within a range corresponding to health while endeavouring to keep side-effects to a minimum. In monitoring tasks, people have to determine whether characteristics of the data stream indicate that some action must be taken or that some previous action has been effective. For example, farmers monitor soil conditions to ascertain whether additional irrigation is necessary. Finally, in forecasting tasks, people judge what is going to happen in the future on the basis of their knowledge of what has gone before. Surveys have shown that forecasting using judgment alone is still very common, particularly in small- and medium-sized businesses. We still do not know a lot about the cognitive processes underlying people’s performance in these tasks. I shall discuss some work on control and monitoring but I shall focus primarily on forecasting. I shall consider various ‘biases’ that have been discovered (trend damping, elevation biases, exaggeration of sequential dependence) and describe some experiments designed to determine why they occur.