Teaching presence, human–GenAI pedagogical partnership, and strategic AI-use regulation in cultural design learning: an instructional–relational–regulatory model

Generative artificial intelligence (GenAI) raises questions about how instructional guidance, partnership perceptions, and regulation relate to students' learning experiences. This cross-sectional study examined an instructional-relational-regulatory model integrating teaching presence (TP), human-GenAI pedagogical partnership (HGP), strategic AI-use regulation (REG), GenAI-use engagement (GUE), cognitive load, and perceived cultural intelligence (CQ). Following a 14-day cultural field work and design activity in mainland China, the primary analysis included 912 students from 21 classes after excluding 32 straightlining responses. A 63-indicator higher-order structural equation model used WLSMV estimation and 2,000 department-stratified whole-class bootstrap draws. TP-HGP showed the largest standardized association (*β* = .695). HGP and REG were positively associated, and both were positively associated with GUE. REG and GUE, but not HGP, had positive conditional associations with perceived CQ. Extraneous cognitive load was negatively associated with REG, GUE, and perceived CQ; intrinsic-load associations were smaller. Results were generally stable across the full 944-case sample, MLR estimation, and leave-one-class-out analyses, although one weak intrinsic-load association was sample-sensitive. Method-component and alternative directional analyses did not eliminate common-method concerns or establish causal ordering. The findings distinguish relational positioning from self-reported regulatory enactment but do not demonstrate cultural-learning gains.