Predicting movement choices in virtual environments

Prediction of landscape preferences using image analysis has been pursued for some time. This paper extends a recently developed approach to image analysis based on explicit determination of pixel-by-pixel depth and content. Another recent experiment monitored path selection choices made in a virtual environment. If the depth and content analysis approach is to be useful for further development of models of human behavior as applied to environmental management, then it needs to be able to predict the choices that people make. The potential of the approach is tested by dividing the path selection experiment participants into three groups on the basis of their choices and deriving regression equations which predict the dominant group choices. An overall prediction success rate of 85% appears promising. Further investigation is required into mechanisms for incorporating the extra information provided by movement in the environment. The importance of separating the population into groups is also questioned.