Extending Automated Scoring to the Figural Form of the Torrance Tests of Creative Thinking

• In Event: Creativity and Assessment

Abstract

The field of gifted education has been moving toward universal screening for gifted identification, but the high cost of assessments remains a significant challenge. Creativity assessment tools like the Torrance Tests of Creative Thinking-Figural (TTCT-F) require manual scoring, which is time-consuming and requires extensive training. In this study, we tested supervised learning techniques with TTCT-F. Using drawings and their titles produced by 464 participants for Activity 1 and 2 in TTCT-F Form A, feature-based classification and fine-tuned computer vision models achieved 80% to 88% accuracy in predicting expert human ratings of originality. Using both text and image features improved the accuracy of the automated scores. These findings show promise for large-scale applications of automated scoring for figural tests of creativity.

Authors

• Selcuk Acar, University of North Texas

• Peter Organisciak, University of Denver

• Denis Dumas, University of Georgia