Assessing Confidence in AI-Assisted Grading of Physics Exams through Psychometrics: An Exploratory Study

TL;DR

This paper investigates the use of psychometric methods, especially Item Response Theory, to evaluate and improve AI-assisted grading of physics exams, aiming to reduce workload while maintaining grading accuracy.

Contribution

It introduces a psychometric framework for assessing AI grading reliability and demonstrates how threshold adjustments can optimize AI performance with minimal human oversight.

Findings

AI achieves R^2 of approximately 0.91 with half the grading load.

AI achieves R^2 of approximately 0.96 with one-fifth of the grading load.

Human oversight remains crucial for uncertain or complex cases.

Abstract

This study explores the use of artificial intelligence in grading high-stakes physics exams, emphasizing the application of psychometric methods, particularly Item Response Theory (IRT), to evaluate the reliability of AI-assisted grading. We examine how grading rubrics can be iteratively refined and how threshold parameters can determine when AI-generated grades are reliable versus when human intervention is necessary. By adjusting thresholds for correctness measures and uncertainty, AI can grade with high precision, significantly reducing grading workloads while maintaining accuracy. Our findings show that AI can achieve a coefficient of determination of $R^2\approx 0.91$ when handling half of the grading load, and $R^2 \approx 0.96$ for one-fifth of the load. These results demonstrate AI's potential to assist in grading large-scale assessments, reducing both human effort and associated costs. However, the study underscores the importance of human oversight in cases of uncertainty or complex problem-solving, ensuring the integrity of the grading process.