The digital supervisor: key to access or shortcutting research

Abstract

Postgraduate students in South Africa and other developing nations face substantial hurdles in completing their research, despite efforts to boost research output and garner subsidies from the Department of Higher Education and Training (DHET). Key issues include research capacity development and supervision burdens. The potential of conversational AIs, like ChatGPT, as research assistants, has been discussed, but more research needs to be focused on using ChatGPT to support novice and student researchers, especially within resource-poor Global South contexts. Large language models (LLMs) such as ChatGPT can support the scientific research process, assisting in generating research questions, developing methodology, creating experiments, analysing data, and writing manuscripts. Responsible use of LLMs in research is crucial, underscoring the need to balance LLM benefits and limitations, retain essential academic skills, and ensure research equity. Our study on ChatGPT's role in postgraduate education offers insights into these areas. We use a Case study approach at a university of technology in South Africa. The research has two main objectives: firstly, it presents a survey examining current ChatGPT usage among postgraduate students, exploring frequency and common applications, and assessing perceived utility in research work. Secondly, it analyses a Design student's experience using ChatGPT to transform initial ideas into a research proposal to understand its potential and constraints as a 'digital supervisor'. Findings underscore the potential of AI tools to boost academic productivity and streamline the research process. It also outlines limitations of AI tools, such as accuracy, potential over-reliance, and creativity concerns. The study highlights the necessity of a balanced pedagogical approach, integrating AI and traditional methods, and promoting ethical AI usage. It underscores challenges in AI tool deployment, like access issues and language barriers, particularly in the South African context. A significant finding is the potential role of AI tools like ChatGPT as digital supervisors, alleviating the burden on human supervisors and strengthening postgraduate culture in design. The study warns against viewing AI tools as complete substitutes for human supervisors and emphasises students' comprehension of AI functions and ethical implications. The research contributes to design education by demonstrating the potential benefits and limitations of using LLMs as 'digital supervisors' to enhance access to postgraduate studies, particularly in the Global South. It emphasises the importance of incorporating LLMs into student learning and research in design responsibly, ensuring students develop requisite research skills and knowledge. Finally, we ask what the implications might be for the field of design is whether there is a need to cultivate our own Design AIs.