Connectionist plan reminding in a hybrid planning model

That planning comprises an automatic (parallel) plan reminding process and conscious (sequential) plan manipulation process and hence is naturally suited for modeling by hybrid connectionist and production system models is the central hypothesis of the author´s research. This hybrid model of planning, which combines symbolic and subsymbolic processing, is different from the traditional artificial intelligence planning models, which use symbolic processing uniformly at all levels. The automatic-plan-reminding process and a connectionist model of the same are described. Plan reminding is defined as the process by which reasonably constrained plans for the given goal or set of goals become automatically available for further conscious processing. It is characterized by dramatic contextual effect, constraint satisfaction and automatic inferencing. The working of the plan-reminding process has been demonstrated using connectionist simulation for a set of errand planning examples. The connectionist network captures some of the errors people make in errand planning.<>