Wednesday 26 March 2025
As educators and researchers continue to explore ways to improve learning outcomes, a new study suggests that artificial intelligence (AI) may be able to provide valuable feedback on student experimentation protocols. The study, which compared AI-generated feedback to that of human teachers and science education experts, found that the AI system was capable of producing high-quality feedback in most areas.
The researchers used a large language model (LLM) to generate feedback on students’ written explanations of scientific experiments. The LLM was trained on a dataset of experiment protocols and was designed to provide feedback on clarity, accuracy, and relevance. The study found that the AI-generated feedback was comparable to that of human teachers and experts in most areas, including providing actionable guidance and offering constructive criticism.
One area where the AI system fell short was in identifying and explaining specific errors within the student’s work context. This is likely due to the limitations of the LLM’s training data, which may not have included examples of this type of error. However, the study suggests that combining human feedback with AI-generated feedback could be a promising approach for enhancing educational practices.
The use of AI in education is not new, but recent advances in natural language processing and machine learning have made it possible to develop more sophisticated AI systems that can provide personalized feedback to students. This study demonstrates the potential of AI-generated feedback as a tool for improving student outcomes, particularly in subjects such as science where experimentation and problem-solving are critical.
The researchers used a dataset of 120 experiment protocols written by high school students to train the LLM. They then asked human teachers and experts to evaluate the same protocols using a set of standardized criteria. The AI system was able to generate feedback on each protocol, which was compared to the human evaluations.
The study found that the AI-generated feedback was most effective in providing guidance on what steps to take next, as well as offering constructive criticism on areas for improvement. However, it struggled with identifying and explaining specific errors within the student’s work context. This is likely due to the limitations of the LLM’s training data, which may not have included examples of this type of error.
The researchers suggest that combining human feedback with AI-generated feedback could be a promising approach for enhancing educational practices. By leveraging the strengths of both AI and human experts, educators may be able to provide students with more personalized and effective feedback.
Cite this article: “AI-Generated Feedback Shows Promise in Improving Student Outcomes”, The Science Archive, 2025.
Artificial Intelligence, Education, Feedback, Students, Experimentation, Protocols, Science, Learning Outcomes, Machine Learning, Natural Language Processing
