Understanding and Fostering Spatial Competence
Sponsor: National Science Foundation, Learning and Intelligent Systems Program
Principal Investigators: Janellen Huttenlocher, University of Chicago; Dedre Gentner, Northwestern University; Nora S. Newcombe, Temple University
Project Summary: This project is concerned with the characterization of spatial competence in intelligent systems. Spatial competence involves representing and reasoning about distance, shape, order, frames of reference, and other relations involving two and three dimensional extent, as well as using diagrams, models, and natural language to communicate such information. Spatial competence is central to many disciplines and real-world skills. Fields such as geography and astronomy and tasks such as navigation and map use involve spatial relations of objects in the world. Fields such as mechanics and organic chemistry and tasks such as assembling and troubleshooting machinery involve understanding spatial relations of parts of objects and their dynamic interplay. In addition, spatial representations are involved in high-level mathematical understanding, and in understanding the diagrams and data visualizations commonly used in many disciplines.
Spatial competence is a fundamental aspect of intelligence, as identified at the behavioral, computational and biological levels. At the behavioral level, both experimental and psychometric evidence have led to the identification of distinct spatial representations and thought processes. At the computational level, recent work indicates that successful spatial reasoning by machine, like human spatial reasoning, requires metric representations that are distinct from the qualitative/propositional information sufficient for nonspatial problems. At the biological level, spatial functioning is known to involve distinct brain areas (i.e., the hippocampus, parietal cortex, and areas of prefrontal cortex).
Spatial competence emerges gradually over long time periods in interaction with environmental input. In an era of rapid growth of technology, a trained workforce requires higher levels of spatial skill than ever before, and it is important to ascertain how to maximize potential in this area. While biological factors may affect spatial skill levels in individuals, there is also evidence that spatial training can contribute substantially to spatial ability. Thus, the proper design of educational input could create higher levels of spatial skill in the population.
Current understanding of spatial competence at the behavioral, biological, and computational levels of analysis is sufficiently advanced to justify optimism that research in the next few years can advance our understanding of the growth of spatial intelligence, providing a firmer basis for improving spatial skills in the population, simulating spatial processes by machine and designing user-friendly interfaces. This work will have both basic and applied significance.
Papers produced with support from this project
Ferguson, R.W. and Forbus, K.D. (1999). GeoRep:
A Flexible Tool for Spatial Representation of Line Drawings. Proceedings of the Qualitative Reasoning Workshop. Loch Awe, Scotland.
Back to Sketching: Towards a computational model page | Back to Projects page | Back to QRG Home Page