Navigating the Dual-View Phenomenon: Social Ambivalence, Ambivalence Literacy, and Lecturer Role Transformation in AI-Integrated Transnational STEM Education

Generative AI chatbots are becoming routine study companions in STEM, which raises a pedagogical question: what do students expect human lecturers to do differently when AI support is ubiquitous? This study examines STEM undergraduates’ expectations for a transformation of the lecturer role and their social ambivalence toward AI chatbots in Sino-foreign transnational education (TNE) programmes in China. We administered an online survey to 467 consenting undergraduates across four partnership institutions (three with sufficient subgroup sizes for institutional comparison). The survey instrument captured adoption readiness, perceived AI-enabled learning enhancement, expected changes to the lecturer role (multi-select), perceived social enhancement and social reduction mechanisms, and perceived support needs; it also asked an open-ended question, collecting 454 usable comments. We report descriptive statistics, χ2 tests, Spearman correlations, and exploratory content analysis results. Students expected lecturers to shift from content delivery to facilitation: 52.7% anticipated that chatbots would handle routine questions, enabling more discussion and practical activities, and 49.7% expected greater emphasis on guiding deep thinking and problem solving. Perceived social impacts were strongly ambivalent: 92.2% endorsed at least one social enhancement and at least one social reduction mechanism, and enhancement and reduction indices were positively associated (ρ = 0.547, p < 0.001), a pattern that remained stable under alternative scoring and response-style trimming (ρ range = 0.526–0.590). Importantly, higher social ambivalence was linked to stronger expectations of lecturer governance and orchestration, including the curation of chatbot resources (42.5% vs. 9.7% in high vs. low ambivalence; χ2(1) = 44.12, p < 0.001) and accuracy checking (27.6% vs. 13.4%; χ2(1) = 8.82, p = 0.003). We therefore propose ambivalence literacy as a conceptual framework for responsible AI integration: a teachable capability to recognise and navigate simultaneous social benefits and risks of AI use, and to translate that recognition into concrete expectations for lecturer governance, orchestration, and facilitative teaching design in AI-integrated transnational STEM programmes.