UM scholar publishes correspondence in Nature to address apprenticeship in AI era

MACAU, June 4 - Tony Zhang Huiquan, associate professor in the Department of Sociology of the Faculty of Social Sciences at the University of Macau (UM), and his collaborator have published a correspondence in the leading international academic journal Nature. Drawing on their experience of using agentic artificial intelligence (AI) tools such as OpenClaw, the authors highlight a pressing challenge facing the scientific community: while AI has dramatically increased research productivity, it is also undermining a core component of education—apprenticeship.

Nature is one of the world’s most influential scientific journals. It publishes about three to four pieces of correspondence per issue, most of which address current trends and issues in academia, research, and education, alongside timely discussions and commentaries. The publication of this correspondence reflects UM’s growing academic strength and international influence, as well as its ongoing engagement with emerging trends in scientific research.

In the correspondence titled ‘AI agents in research: when productivity comes at the cost of apprenticeship’, the authors point out that students’ research capabilities rely heavily on hands-on guidance from mentors. Students have to engage in active practice and make mistakes and corrections under the guidance of mentors in order to acquire basic research skills, tacit knowledge, and disciplinary judgment. However, as scientists increasingly integrate AI agents into their research workflows, junior researchers are being replaced in many basic research tasks that once fell to them. With fewer research tasks, less frequent group meetings, and weakened apprenticeship, students are gradually being cut off from the critical processes involving decision-making, weighing up options, troubleshooting errors, refining methods, and calibrating instruments. As a result, students are missing out on the very learning opportunities that provide the essential ‘cognitive scaffolding’ young scholars need to develop. When senior researchers use AI tools to accelerate their output, they may inadvertently deprive students of the opportunity to learn and grow.

The correspondence emphasises that scientists are responsible not only for advancing research, but also for mentoring students. The authors call for a balance between research productivity and the essence of education. They suggest the following practices: integrating AI into research training and working alongside students to master new tools; preserving essential ‘learning by doing’ opportunities while improving research productivity; positioning AI as a shared assistant for both mentors and students, rather than a replacement for students; and ensuring that technology empowers collaboration between mentors and students without compromising the long-term goals of education. The correspondence addresses a critical question at the heart of the shift triggered by generative and agentic AI: How can we protect the apprenticeship model that has long been central to scientific education while enhancing research productivity?