AI-ASSISTED GRADING IN HIGHEREDUCATION: ACCURACY, EFFICIANCY, AND FAIRNESS COMPARED TO HUMAN RATERS

This study investigates the use of artificial intelligence (AI) in higher education essay grading, emphasizing justice, validity, efficiency, and dependability. Modern AI systems are made to deliver assessments more quickly and consistently, whereas traditional hand grading is frequently laborious and subjective. Three hundred undergraduate essays were compared and evaluated by an AI system that was in line with the rubric and by qualified human raters. According to the results, AI grading performed close to human-human agreement (0.84), with a Quadratic Weighted Kappa (QWK) of 0.81 in comparison to human raters' consensus scores. Furthermore, AI shortened the grading process by almost 70%, demonstrating its effectiveness in extensive evaluation. Nevertheless, a fairness investigation showed some biases, with AI giving native speakers slightly higher marks. These results imply that AI-based grading can efficiently enable consistent, fast, and extensive evaluation in higher education when paired with teacher supervision and fairness checks.