Student Essays and Human and LLM-generated Reviews - Academic Writing Study

Anonymized data and code for a study examining university student impressions of human- and AI-written academic reviews. 

Each student received two reviews for an essay written for an academic writing course, one of which was written by the instructor and one by an LLM. 

Data was collected following initial reading and two weeks later after engaging with the reviews (preparing a response, making changes, etc.). Students were significantly better at recognizing LLM-generated material after engaging with the reviews, though still at random odds.