Seedlabs
Source research

Generative AI tool use enhances academic achievement in sustainable education through shared metacognition and cognitive offloading among preservice teachers

Javed Iqbal, Zarqa Farooq Hashmi, Muhammad Zaheer Asghar, Muhammad Naseem Abid · 2025 · 67 citationsRead the paper

The integration of generative artificial intelligence tools in education has emerged as a transformative approach to enhancing learning outcomes, particularly in the context of sustainable development goals (SDG4). Therefore, the present study investigates the connection between generative artificial intelligence tool usage (GenAITU) and academic achievement (AA) in the context of SDG4. We assessed the mediating role of shared metacognition (SMC) and cognitive offloading (COL) in this relationship among preservice teachers (PSTs). The indicators, including performance expectancy (PE), effort expectancy (EE), facilitating conditions (FC), and use behavior (UB), are derived from adapting the Unified Theory of Acceptance and Use of Technology 2 (UTAUT2) for GenAITU. The authors surveyed 465 students from five universities in Wuhan, China, using a 7-point Likert scale through a time-lag design. Statistical analysis was performed through partial least squares structural equation modeling (PLS-SEM), to determine the relationship between variables. Findings indicated that two components of GenAITU, namely PE and UB, showed significant positive associations with AA, while the other two, EE and FC, did not show significant and positive relationships with AA. Results also showed that three dimensions of GenAITU, namely EE, FC, and UB have a positive and significant association with SMC while PE has a positive and significant connection with SMC. All four components of GenAITU like PE, EE, FC, and UB have positive and significant links with COL. SMC and COL have a positive and significant relationship with AA. Results also indicated that SMC mediated the connections between GenAITU (EE, FC, and UB) and AA. Outcomes also indicated that COL mediated the connections between GenAITU (PE, EE, FC, and UB) and AA. The current study shows that SMC and COL were strong mediators of the association between GenAITU and AA. The results of our study provide guidance to teachers, curriculum planners, and university management to successfully integrate GenAITU into the education for PSTs.

What the full text shows

Shown in the lab
  • The model explains the variance in academic achievement (AA), cognitive offloading (COL), and shared metacognition (SMC). — 40.8% variance in AA, 34.5% in COL, and 30.7% in SMC (Preservice teachers (PSTs) in the context of SDG4)
  • Performance expectancy (PE) and use behavior (UB) have significant positive associations with academic achievement. (Preservice teachers (PSTs))
  • Cognitive offloading (COL) has a more substantial contribution to the variance in academic achievement than shared metacognition (SMC). — f-square of 0.127 for COL vs 0.049 for SMC (Preservice teachers (PSTs))
  • Effort expectancy (EE), facilitating conditions (FC), and use behavior (UB) positively and significantly contribute to shared metacognition. (Preservice teachers (PSTs))
  • All four GenAI tool usage components (PE, EE, FC, and UB) positively and significantly contribute to cognitive offloading. (Preservice teachers (PSTs))
  • Shared metacognition and cognitive offloading both have positive and significant relationships with academic achievement. (Preservice teachers (PSTs))
Setup
465 usable responses from bachelor's, master's, and doctoral students across five universities in Wuhan, China; two-phase survey conducted 2 weeks apart.
How
The study used a quantitative time-lag survey design and analyzed data using partial least squares structural equation modeling (PLS-SEM) via SmartPLS 4. It adapted the UTAUT2 model to measure GenAI tool usage and assessed the mediating roles of shared metacognition and cognitive offloading.
Limitations
Sample limited to China, where some AI tools are restricted · Reliance on self-reported data introducing potential social desirability bias · Time-lag design does not fully establish causal relationships · Lack of behavioral data (e.g., system usage logs) for triangulation

Summarized by Gemma 4 from the open-access full text · 7 Oct 2026. Read it