PROBLEM SOLVING ON A STOCHASTIC PROCESS

This study is concerned with man-computer cooperation in which the computer initiates decisions and the man monitors and can alter these. The task used involved controlling an information process by making decisions at each of a sequence of points. The problem of control was to find an optimal procedure that jointly minimized two variables. The solutions of subjects developed over a series of trials were compared with that of an optimal programme under different conditions. Subject performance was nearly always inferior to that of the programme, it was impaired by increasing the rate of the information input and failed to benefit from reduction of uncertainty in the input. Giving trial knowledge of results helped convergence towards an optimal solution. Subjects who had gained experience in the task were given computer solutions to monitor. They degraded optimal solutions and improved inferior solutions towards their own level of performance and failed to benefit from this experience in subsequent tests.