Improved Latent Semantic Analysis Approach in Automated Scoring of Rater-Mediated Assessments: Example of Creativity Assessments

• In Event: Thursday Roundtable Session 9:00 am
  In Roundtable Session: AI, Psychometrics, and Data Analysis (Table 24)

Abstract

As a rater-mediated performance assessment, subjective creativity assessment (SCA) relies on the rater’s judgment of creativity. Although automated scoring models using Latent Semantic Analysis (LSA) have been developed for SCA during the past few decades, they are mainly designed for scoring the originality of divergent thinking (DT) ideas. Creative solutions that individuals generate to solve a real-world problem are sample-specific and more complex but have a smaller sample size than DT ideas, which pose challenges to traditional LSA methods. Therefore, we propose an improved LSA approach to automated scoring for the originality of creative solutions and demonstrate it with data collected from 6th-graders solving a science problem. The significance of the study was discussed.

Authors

• Ziyue Wang, University of Science and Technology of China

• Haiying Long, University of Kansas

• Jue Wang, University of Science and Technology of China