14/09/2026
Becoming Familiar with the Algorithm: Generative AI and the Research Proposal Problem of a South African Master’s Student
Thandiwe Nkosi was sitting at her desk in a simple apartment in Johannesburg, the blue light of her laptop reflecting a partly completed research proposal for her Master’s in Education at a private higher education institution. Language policy and digital access in township secondary schools was the topic, demanding a concise conceptual framework and a defendable literature review. Deadlines were closing in. There had been no feedback from her supervisor. One night, Thandiwe decided to open ChatGPT and typed in a carefully crafted prompt: “Provide a conceptual framework for language policy and digital inequality in South African township schools by applying postcolonial theory.”
In a split second, the tool provided coherent paragraphs, important theorists and even sample research questions. Relief surged through Thandiwe; the summary made sense out of dense journal articles which she had struggled with after teaching all day. She copied text, started paraphrasing. Then doubts began to cloud her mind. Where were the references? Did the linkages hold up upon closer examination? Would her supervisor detect the stylistic signatures of the machine?
Her dilemma mirrors what is happening amongst postgraduate students across Africa and beyond beyond. According to Maphosa, Khoza, Tlomatsana and Pitjo (2026), a survey conducted in a mixed-methods approach among 75 postgraduate diploma, honours and master’s students studying at a private higher education institution, almost three-quarters of them are using generative AI tools such as ChatGPT for research ideas generation, article summaries and clarification of difficult concepts. These students had positive attitudes towards the support. However, more than a quarter of them refrained from using the tools due to the fear of plagiarism, algorithmic bias, privacy concerns and the belief that they would hamper their cognitive development. More than 78 percent were familiar with the policy of their institution on the use of the generative AI tools but the uncertainty about the responsible usage was widespread.
Thandiwe knew the pros and cons of the tool. It was speeding up her work and reducing her isolation. It could be making up the references, depriving her of critical engagement with the South African scholarship and transferring ownership of her intellectual work. The guidelines of her institution on generative AI were clear but vague concerning implementation issues.
She closed the chat window and looked at the blank page on her laptop. Should she keep using the tool under very strict conditions, checking every citation in peer-reviewed journals, paraphrasing every statement in her own language, declaring her usage of the tool to her supervisor, or give it up and risk failing to submit the research proposal in time due to its insufficient quality?
What do you think Thandiwe was supposed to do? What are your own struggles with AI?
Reference: Maphosa, M., Khoza, L., Tlomatsana, C. and Pitjo, W., 2025, November. Exploring postgraduate students’ use and perceptions of generative AI as a research support tool: A South African case. In Southern African Conference for Artificial Intelligence Research (pp. 437-451). Cham: Springer Nature Switzerland.
'sstudent