Postgraduate Students’ Judgements of GenAI Use in Research: A Scenario-Based Study Across Four African Countries
Abstract
Generative artificial intelligence (GenAI) is increasingly used in postgraduate research, while recent scholarship shows that students often navigate a grey area between legitimate assistance and substitution of their own intellectual work. Less is known about how postgraduate researchers judge graded forms of GenAI involvement across institutional contexts. This multi-site scenario-based study examined 233 survey submissions from five institutional sites in Botswana, Senegal, Tanzania, and Uganda. Participants evaluated eight postgraduate-research scenarios on two five-point outcomes: whether the conduct was unethical and whether it would be against institutional rules. The scenarios varied the amount of AI-generated material incorporated into the research product, demonstrated understanding, and stage of use. The study was framed by Jones’s issue-contingent perspective and Treviño’s person-situation interactionist model. Amount of AI-generated material was by far the strongest pooled factor for ethicality (F = 829.400, p < .001, partial η² = .316) and perceived rule violation (F = 843.205, p < .001, partial η² = .323), and it was the only positive significant scenario predictor in every country-specific ordinal model for both outcomes. Understanding was much weaker and stage of use was negligible overall. Cross-country differences were concentrated in particular scenarios rather than forming a stable ranking. The study extends scenario-based GenAI ethics research by quantifying which features of ambiguous postgraduate-research situations most strongly shape ethical and institutional judgements.
