R1-[ImageSpace] - Overview

Mental Representations of Spatial Environments

 

Mental reasoning about spatial environments is often based on spatio-analogical or quasi-pictorial mental representation structures. As human working memory for spatio-analogical knowledge processing is severely restricted in capacity, mental processes dynamically construct and explore task-sensitive representations to obtain a desired piece of spatial information. We develop the computational cognitive architecture Casimir to model the construction and inspection of mental representations of spatial environments and to explore these models from a computational perspective. The project is inherently interdisciplinary, combining modeling techniques from informatics with empirical psychological studies. Experimental results are employed for both informing and evaluating the modeling work. At the same time the developed models guide experimental work, creating a mutually informing modeling-experimentation cycle.

The project focuses on designing and implementing a processing architecture based on spatio-analogical and diagrammatic structures that provides the basis for a computational description of the corresponding mental processes. In doing so, computational cognitive models of long-term memory, working memory (e.g., spatial mental models, visual mental images), and control mechanisms (e.g., control of spatial reference frames) have been developed. These models have been found to closely mirror human behavior in a wide range of spatial abilities such as spatial reasoning, mental image re-interpretation, perspective taking, and spatial language use, among others. The resulting models will be applied in prototypical spatial task assistance systems that complement internal representations by external representations interactively to compensate mental processing restrictions.