Two frameworks to guide discussions around levels of acceptable use of generative AI in student academic research and writing
The capacity of generative AI tools, such as ChatGPT, to perform at human-like levels on a range of tasks has created a challenging pedagogical tension within education. On the one hand, such tools pose significant threats to academic integrity, but on the other hand, employers will also expect graduating students to be proficient in the use of such tools so as to improve their productivity and performance levels. How best to balance this tension and produce students who can both work proficiently with discipline-relevant generative AI tools and also add significant value to anything a generative AI system can produce will be a long-term research project for educators. In the short term, however, since many students are already widely using or exploring the use of such tools, it is important for educators to clearly delineate for students what level of use is acceptable or unacceptable for each assessment task. To help educators frame any discussion with their students about levels of acceptable use, this paper proposes two frameworks for the uses of generative AI to support academic writing. The first framework is relatively simple, and generalises ideas from long term discussions around the support-collaboration-collusion continuum to the case of generative AI support for academic writing. The second proposed framework is two dimensional, considering the different levels of support possible at each of six stages in the academic writing process. In addition, since helping students learn effective “prompt engineering” will need to be a pedagogical goal, use examples at each level of the second framework are provided, together with commentary about effective prompt engineering, issues to watch out for, and possible ways to acknowledge any such use. It is hoped that the analyses and examples provided in this paper will provide useful foundations for both discussions between educators and their students, and future research into how best to integrate the use of generative AI tools in higher education pedagogy. Some suggestions for the needed research are also given. Since similar issues as those discussed above arise for research students when writing for publication, some discussion of these issues in relation to (evolving) academic journal policies in this area is also provided.