Enhancing Automated Short-Answer Grading: The Impact of a Robust Scoring Rubric

• In Event: Applications of AI II

Abstract

We examined the effect of a robust scoring rubric to mitigate scoring inconsistency on an automated short-answer grading system. Human scoring inconsistencies may have adverse impact on performance, reliability, and validity. We refined a Transformer-based model using the new scores generated from the robust rubric to improve the scoring performance.

Authors

• Wallace Nascimento Pinto, University of Florida

• Lodi Lipien, Florida Virtual School

• April Fleetwood, Florida Virtual School

• Jinnie Shin