A comparison of block-stacking heuristics used by preschool children and classic artificial intelligence planning paradigms

Despite the large body of research in Psychology concerning human planning and problem-solving both by adults and children, the study of planning and problem-solving in the field of Artificial Intelligence has proceeded along its own development with very little concurrent exploration of the methods people use to plan and solve problems. Some of the classic planning programs are unable to solve problems which are trivial for children, and it may be that by exploration of the methods children use we will discover certain heuristics which can be incorporated into AI planning paradigms. This thesis explores this possibility. Children aged 3 to 5 years were recorded performing a block-stacking task which simulates the type of problem given to planners to test their efficiency. The data were analyzed in order to determine those heuristics which are common to planners and children as well as those which are unique to the children. Based on this analysis, the psychological validity of the planning programs are evaluated.