THE DESIGN AND INTERPRETATION OF HUMAN CONTROL EXPERIMENTS

Abstract

PART I is a general statement of some of the factors which must be taken into account in the design and interpretation of human control experiments. In particular the non-ergodic character of the asymptotic learning state (statistical quasi-equilibrium) demands special precautions and a large amount of data. Attention is drawn to the part played by bias and noise.

PART II is an account of certain experiments with a machine incorporating computer elements so that statistical data may be accumulated rapidly, interalia in the form of mean square deviations. The experiments described relate to the various phases of learning and the regression of mean square deviation on sensitivity in the ‘ asymptotic ’ state.