Colloquium Details
Look without a homunculus: Incorporating bottom-up processes in modeling complex visual search in ACT-R
| Author: | Duncan Brumby Cardiff University |
|---|---|
| Date: | August 19, 2004 |
| Time: | 15:30 |
| Location: | 220 Deschutes |
Abstract
Visual search is an important component of human information interaction. The development of information visualizations, such as the degree-of-interest (DOI) tree, presents the user with a complex visual scene to navigate and search. Previously, cognitive models have been developed within the ACT-R cognitive architecture of simple, routine menu search. Model development of visual search in a complex visual scene, such as the DOI tree, or other information visualizations, reveals important limitations of ACT-R's perceptual/motor theory. First, the modeler acts as a homunculus to the system. Second, the simplified account may not be able to predict the influence of bottom-up interactions on the guiding of visual attention. One solution may be to implement Logan's (1996) CODE-TVA theory within the ACT-R framework. In addition, the theory leads to a parsimonious explanation of recent data within the literature, which is not well explained by current models.
This talk pertains to work in progress with Stu Card and Peter Pirolli at Parc (formerly Xerox Parc).
Biography
Duncan Brumby, a Ph.D. student in Psychology from Cardiff University in Wales.