Adaptive spatial planning: An ACT-R model

The paper presents an experiment on spatial planning in 2D maze-like environments in which participants had to build and execute on-line a plan to reach a target. The experiment manipulated three factors: type of instructions (two levels), number of obstacles in the mazes (four levels), and length of the minimum path to the target (three levels). The instructions were conceived to vary the importance of the two main goals of the decision behavior: i.e., accuracy maximization and effort minimization, respectively. The results support the hypothesis of the adoption by participants of two different solution strategies as an effective response to the instructions. The strategies differ in the amount of time devoted to execution vs. reasoning, and in the way these processes are interleaved. We discuss the implications of the results for the adaptive view of cognition, and propose an ACT-R model to simulate the participants’ strategic behavior. The model explains the differences in the results by assuming that participants, aiming at maximizing accuracy vs. minimizing effort, differ in the likelihood and in the location of the intermediate goals they set in their solution paths. In the conclusions, we analyze the contribution of ACT-R simulation as a tool for studying the adaptive spatial planning.