Can Large Language Models Make the Grade? An Empirical Study Evaluating LLMs Ability to Mark Short Answer Questions in K-12 Education

This paper presents reports on a series of experiments with a novel dataset evaluating how well Large Language Models (LLMs) can mark (i.e. grade) open text responses to short answer questions, Specifically, we explore how well different combinations of GPT version and prompt engineering strategies performed at marking real student answers to short answer across different domain areas (Science and History) and grade-levels (spanning ages 5-16) using a new, never-used-before dataset from Carousel, a quizzing platform. We found that GPT-4, with basic few-shot prompting performed well (Kappa, 0.70) and, importantly, very close to human-level performance (0.75). This research builds on prior findings that GPT-4 could reliably score short answer reading comprehension questions at a performance-level very close to that of expert human raters. The proximity to human-level performance, across a variety of subjects and grade levels suggests that LLMs could be a valuable tool for supporting low-stakes formative assessment tasks in K-12 education and has important implications for real-world education delivery.

Introduction. Assessment and feedback are crucial components of the learning process but are resource intensive to do regularly at scale [4, 21]. Unlike the high-pressure environment of traditional high stakes examinations, formative assessments are generally intended to be diagnostic, enabling students and teachers to adapt their approach within or in-between lessons to maximize learning. Also sometimes referred to as assessment for learning, formative assessment has been shown to lead to significant improvements in learning outcomes [12], however, scaling formative assessment practices have traditionally proven to be challenging due to the significant costs and logistical demands involved [4]. Closed-response assessment questions, such as multiplechoice and true/false, are commonly used in formative assessment, and have the benefit of being efficient to grade [18]. However, they have several drawbacks, such as the possibility of students relying on test-taking strategies, a potential lack of face validity, and the complexity in generating multiple answer options [2, 16].

Discussion / Conclusion. In this experiment, we aimed to explore how well GPT-4 performs at marking open text responses to short answer questions and how performance varies across domain areas, grade levels, and question difficulty. Overall, we found that GPT-4 performed well human-level performance (Kappa, 0.75). That GPT-4 performs in line with expert human raters is consistent with prior work which tested model performance at scoring short-answer reading comprehension questions [13]. We found only very modest variation in model performance based on subject, gradelevel, and question difficulty. Our results from this experiment with Carousel build on our prior work focused on reading comprehension and, importantly, suggest that LLMs like GPT-4 could potentially be used for a variety of low-stakes assessment tasks across different domain areas and grades.