Key points are not available for this paper at this time.
Large language models (LLMs) such as ChatGPT are being used for outlining, writing assistance, and higher-level research tasks in many areas of science. The purpose of the work described here was to assess ChatGPT version 3.5's ability to help upper-level Analytical and Inorganic Chemistry students find methods and green modifications to use in laboratory proposals. Some students found that ChatGPT provided a helpful starting point, but it mostly gave vague and erroneous responses when asked to provide upper-level methods in these areas, particularly in green chemistry. Few ChatGPT suggestions were useful for students' proposals. Students usually used reasonable prompts, and they often evaluated ChatGPT outputs correctly; however, most students reported they were not confident using its responses. Some lacked confidence in ChatGPT outputs because it rarely provided references, and some said they preferred formal literature searching. Despite ChatGPT's deficiencies in producing usable methods, we found evidence that these were valuable exercises for helping students think critically about procedures and artificial intelligence (AI) outputs. An accompanying anonymous survey suggested that student opinions of ChatGPT and AI technologies varied depending on what they had used them for. Based on this work, we offer suggestions for other instructors who wish to design similar method development assignments that incorporate LLMs.
Ruff et al. (Tue,) studied this question.