Unlocking the Black Box: Use Bayesian Knowledge Tracing to Evaluate Perceptual Scaffolding in Math Intervention

• In Event: AERA Roundtable Session Sunday 7:45 am - Gold 4
  In Roundtable Session: Cognitive Processes, Human–Technology Interaction, and Learning Analytics

Abstract

Do perceptual supports act as scaffolds or crutches in educational practice? The current study employed a learning analytics approach to estimate how perceptual cues bias immediate problem-solving and affect the acquisition of underlying skills. We fit a Bayesian Knowledge Tracing model to a math intervention dataset comprising problem-solving accuracy from 688 U.S. middle schoolers. Perceptual cues were embedded in intervention materials and randomly assigned across participants. Congruent perceptual cues increased the guessing rate and reduced the slipping rate, indicating a perceptual grouping effect. Nevertheless, incongruent color cues revealed the highest learning rate and the lowest forgetting rate, functioning as desirable difficulties. Our study presents an example of using learning analytics technologies to examine math cognition theories and inform instructional design.

Authors

• Puyuan Zhang, Worcester Polytechnic Institute

• Siddhartha Pradhan, Worcester Polytechnic Institute

• Erin R Ottmar, Worcester Polytechnic Institute