Uta Priss
u.priss@ostfalia.de
Informatik, Ostfalia University, Wolfenbüttel, Germany
Abstract
A homework exercise in my introductory mathematics class for first year computing students consists of assigning a simple set or number theoretical statement to each student to prove and implement for some examples in Python. Because logic, but not specifically proof techniques, are taught in the class, in previous semesters the students were encouraged to consult textbooks or the internet for the proofs. This semester I decided to require students to use ChatGPT for the exercise and to let them reflect on its usefulness. In the second part of the exercise each student reviews the anonymous submissions of 3 other students. Because ChatGPT is evolving very quickly I found only very few, somewhat similar studies in the literature so far (e.g. Rojas (2024) and Park and Manley (2024)). I observed that ChatGPT appears to have an equalising effect on most of the students’ work apart from the very low end students which seem to invest insufficient time and effort no matter what. Very few students achieved full marks on the first part of the exercise. High performing students may be held back by ChatGPT because it stops them from generating their own solutions and thinking freely. With ChatGPT most students produced reasonably good proofs and code, but usually with some minor problems. The answers from ChatGPT tend to be overly verbose and complicated which somewhat confused the students and made it difficult for them to detect errors. ChatGPT’s Python code is correct and mostly good, but not always exactly what is required (which the students did not always realise). The marks for the second part of the exercise were based on the students’ comments and reflections. It appeared to me that the students overlooked more errors than in previous semesters, maybe because ChatGPT’s answers are well-worded and asserted in a very confident manner. Because the students’ overall performance was not better than in previous semesters, it did not seem to me as if the students received undeservedly good marks or that using ChatGPT was a kind of cheating for them. If accepted, I will provide a more detailed analysis of my observations of how ChatGPT changed the exercise.
Keywords
ChatGPT, mathematics homework exercises, teaching logic and proofs
References
Rojas, A.J.. (2024) An Investigation into ChatGPT’s Application for a Scientific Writing Assignment. Journal of Chemical Education, 101(5), pp.1959-1965.
Park, H. and Manley, E.D. (2024) Using ChatGPT as a proof assistant in a mathematics pathways course. The Mathematical Education, 63(2), pp.139-163.