This item is non-discoverable
Raven's matrices in the clarion cognitive architecture
Loading...
Authors
Mekik, Can, Serif
Issue Date
2022-08
Type
Electronic thesis
Thesis
Thesis
Language
en_US
Keywords
Cognitive science
Alternative Title
Abstract
The Raven's Matrices family of psychometric tests measures fluid intelligence, a broad psychometric ability involved in reasoning, analogy, and induction. The present dissertation introduces Xyrast, a broadly-scoped computational cognitive model of human response processes on this family of tests in the Clarion cognitive architecture. Xyrast's solution process involves an interruptible search (anytime algorithm) for visual patterns that may inform response choice. The model focuses on response strategy, working memory capacity, and persistence as psychological parameters. Distinctively, persistence is modeled through the influence of motivational variables on the response process. Furthermore, the model's pattern-finding ability is powered by a connectionist visual-relational reasoning architecture and, as part of this architecture, the present work introduces a novel perceptual field representation to facilitate object selection. Xyrast simulates response choice, response time, and fixation data. Simulations reveal that the model captures several phenomena reported in the literature including item complexity effects, correlations among eyetracking measures, variations in correlations between item scores and working memory capacity, and a negative effect of response latency on scores that varies according to item difficulty and subject ability. Interestingly, the relationship between performance and basic psychological variables is found to be highly sensitive to contingencies like item structure and response strategy. This finding suggests that the relationship between fluid intelligence and basic psychological parameters may be more dynamic than previously thought.
Description
August2022
School of Humanities, Arts, and Social Sciences
School of Humanities, Arts, and Social Sciences
Full Citation
Publisher
Rensselaer Polytechnic Institute, Troy, NY