A study of human control in a stochastic multistage decision task

Models of decision making can be grouped into two general classes: static and dynamic decision making. The first consists of those tasks where a single decision is made, the subject is told of the results of his decision, and no further application is made. In dynamic decision making, subsequent decisions depend in part on past experience in the task and thus learning is involved in the act. The latter sort of decisions can be further broken down into two types: those that do not affect the environment in which the decision maker is behaving, and those involving the future environment. A decision maker who can actively manipulate the environment by his decisions is conceived of as a controller. This article considers a dynamic programming model for this type of decision‐making task.