Leveraging Generative AI for Assessing Student Reports in Engineering Education

This paper examines the application of Generative AI (henceforth GenAI), specifically OpenAI's advanced Large Language Models (LLMs), in evaluating student reports within challenge-based courses with focus on Sustainability and Ethics (S&E). Traditional grading methods, heavily reliant on manual effort and subject to human biases, present a significant workload for educators and often lack consistency. By integrating LLM into the grading process, this study explores the feasibility of automating assessment tasks, aiming to achieve more objective and efficient evaluation while reducing evaluators' workload. The research method entails preprocessing and anonymizing student reports, promptifying existing rubrics for LLM compatibility, and analyzing the AI-generated assessments against human- graded benchmarks. Preliminary findings suggest that LLMs can complement human grading by providing consistent evaluations under certain conditions, such as when reports are text-based and rubrics are clearly defined. However, limitations such as the model's reasoning capabilities and handling of non-textual information indicate areas for future research. This research contributes to ongoing discussions on the potential of GenAI in education, underscoring the need for further exploration into AI- assisted assessment tools that could enhance the transparency and efficacy of grading practices in engineering education.