Spatial adaptation: modeling a key spatial ability

Abstract

ABSTRACT Spatial adaptation is the process of adjusting one’s mental representations for a task, so that spatial details necessary for performing the task are captured in the representations, whereas irrelevant details are ignored. We believe this process plays a critical role both in spatial ability tests and in STEM domains because it produces problem-tailored representations that can facilitate mental manipulation by representing only task-relevant details. Here, we present a computational model that illustrates the importance of spatial adaptation in a mental rotation task. The model automatically generates shape representations by segmenting objects into parts at concavities. It adjusts its representations in two ways: by varying the number of parts used to represent a shape, and by varying the types of information encoded for each part. Critically, the model can adapt to a mental rotation task by adjusting the degree of detail in its shape representations automatically, based on how much detail is needed to distinguish the shapes from distractors.