“The use of chatbots inmedical education is an emerging trend that is welcomed by many educators and medical professionals. In particular, the use of ChatGPT, a large languagemodel of OpenAI, offers a variety of benefits for students and educators alike [...]” [1]. So far so amazing, the passage already points to the whole dilemma: will teaching at universities ever be the same after ChatGPT as it never was anyways? We had a Cologne term paper in the “field of competence carcinogenesis” (interdisciplinary teaching in the first preclinical study semester) generated in triplicate by ChatGPT, each with identical queries, and received three different two-page texts including literature citations according to APA style. These have been examined by two detector programs (Groover, Writer) to determine whether they were written by a human or a bot. Both programs could not detect them as machine-written (cave: short texts are practically undetectable). The search for plagiarism with the software PlagAware did not reveal any conspicuous passages worthy of consideration (approx. 3-5% agreement with already published texts). The papers were forwarded unchanged to the assessing tutors with the result that two papers were assessed as “passed” and one as “failed”. The poor performance was due to certain terms used in the field of competence that was not named, as well as a non-matching literature citation. What next? Let’s ask ChatGPT: “...If students were able to access ChatGPT and ask questions during the exam, they could theoretically receive answers from ChatGPT that could help them answer exam questions...” [2].
No takes yet. Share an insight, caveat, or question.
Moritz et al. (2023) studied this question.