Artificial intelligence and the future of learning and assessment in agricultural and applied economics

Abstract

Using a repository of historical student responses to an actual course-assigned essay prompt and a series of artificial intelligence (AI)-generated responses to the same prompt, we conduct a single-blind, randomized experiment to evaluate the performance of AI in agricultural and applied economics education. Further, we assess instructors' ability to detect the use of AI. We find that AI-generated responses to the essay received statistically significantly higher scores than those of the average student. Instructors who had previous exposure to dialog-based AI were 13 times more likely to accurately detect AI-generated essays than instructors without previous exposure to the technology.