Pedagogical partnerships with generative AI in higher education: how dual cognitive pathways paradoxically enable transformative learning

Synthesis: Wang and Zhang (2026) use a four-study mixed-methods design across China, Europe, and the United States (N=912) to examine how students' framing of GenAI as a pedagogical partner affects transformative learning through two simultaneous cognitive pathways — vigilance (critical evaluation) and offloading (strategic delegation). Their central, counterintuitive finding is that strategic cognitive offloading, contrary to traditional assumptions, can enhance transformative learning by liberating mental resources for higher-order reflection once it exceeds certain thresholds. Efficiency orientation, rather than undermining learning, acts as an amplifier of both vigilance and offloading. The study reconceptualizes student-AI dynamics from zero-sum competition to synergistic collaboration.

Core Finding

Framing GenAI as a pedagogical partner simultaneously activates two cognitive pathways — critical vigilance and strategic offloading — and, paradoxically, both can contribute to transformative learning. The study rejects the binary view that trust/skepticism and delegation/depth are opposing poles. Partnership orientation triggers vigilance (students critically evaluate AI outputs because they share accountability for co-created work) and, at the same time, legitimizes offloading routine cognitive tasks as collaborative efficiency. Against the classic assumption that delegation erodes learning depth, strategic offloading — above a threshold — frees cognitive capacity for higher-order reflection, enhancing the Metacognition and perspective transformation central to transformative learning. Efficiency orientation is recast from a barrier into an amplifier, strengthening both pathways through pragmatic risk management.

Dual Cognitive Pathways

The theoretical model integrates two cognitive responses available to students interacting with GenAI:

• Cognitive vigilance — systematic critical evaluation of AI content (error detection, reasoning verification, source cross-referencing, epistemic appraisal of whether outputs meet disciplinary standards). Partnership orientations activate this through relational accountability.
• Cognitive offloading — strategic delegation of cognitive tasks (summarization, synthesis, drafting, data analysis) to the AI, grounded in metacognitive frameworks of external cognition. Partnership reframes this from intellectual abdication to strategic resource management.

The findings show both pathways positively associate with transformative learning, resolving the theoretical tension between delegation-as-erosion and engagement-as-enhancement. Efficiency orientation positively moderates the partnership→offloading link and (in this configurational analysis) does not uniformly suppress vigilance.

Methodology

The study uses a rigorous mixed-methods design with four complementary studies: (1) PLS-SEM to test the mediation/moderation architecture; (2) importance–performance map analysis (IPMA) to translate findings into pedagogical priorities; (3) fuzzy-set qualitative comparative analysis (fsQCA) for configurational, equifinal pathways; and (4) semi-structured interviews with 45 business students for process-level triangulation. A three-wave, time-lagged survey across three cultural contexts minimizes common method bias. Constructs (Human-GenAI Pedagogical Partnership, Cognitive Vigilance, Cognitive Offloading, Transformative Learning Experience, Efficiency Orientation) were validated through an expert panel, a 90-respondent pre-test (α>0.700), back-translation, and convergent/discriminant validity checks (HTMT<0.850).

Relevance to the Knowledge Base

This article is a strong theoretical and empirical anchor for the knowledge base's coverage of Human AI Collaboration and the cognitive dynamics of Student-AI Interaction. It engages the Learner Agency, Trust, Metacognition, and Critical Thinking concepts, and offers a constructive counterpoint to articles focused on AI-induced cognitive decline or academic dishonesty. It also informs Educational AI Policy and Learning Design by providing evidence-based guidance on how to integrate GenAI in ways that support rather than erode deep learning.

What this means for practice

• Instructors. Teach students to pair delegation with vigilance: set tasks in which they must verify, cross-reference, or challenge AI output before it counts as their own work, and avoid treating GenAI use as something to be banned or endorsed wholesale.
• Instructors. Add explicit reflection points where students articulate their evolving relationship with the tool, and require self-regulation and critical evaluation rather than output completion alone.
• Administrators. Shift institutional messaging from warnings about AI shortcuts to recognizing thoughtful delegation as cognitive resource management — in this study efficiency orientation amplified both the vigilance and offloading pathways rather than suppressing critical engagement.
• Designers. Write assessment rubrics that reward the sophistication of human-AI interaction, including strategic delegation, critical evaluation, and synergistic knowledge creation, not just the quality of the submitted artifact.

Limitations

• Every construct — partnership orientation, offloading, vigilance, transformative learning, efficiency orientation — was measured through student self-report, so the mediation and moderation results describe perceived rather than demonstrated learning.
• The data come from business-school students (N = 912 valid three-wave responses across China, Europe, and the US) working primarily with text-based LLMs; the authors state generalizability is bounded to that disciplinary context and to text-based models.
• The 45 semi-structured interviews used for process-level triangulation were drawn exclusively from the Chinese subsample, so the qualitative accounts cannot speak to the European or US experience.
• Although the three-wave, time-lagged design temporally separates predictors and outcomes, the authors concede that longer windows and objective performance indicators are still needed to show that partnership routines stabilize and transformative insight endures.

Citation

Wang, S., & Zhang, H. (2026). Pedagogical partnerships with generative AI in higher education: how dual cognitive pathways paradoxically enable transformative learning. International Journal of Educational Technology in Higher Education.