As generative artificial intelligence (GenAI) tools become increasingly integrated into academic settings, questions are mounting about how students are using these tools and whether such use supports or undermines their own learning. One behavior drawing particular attention is the tendency of some students to delegate metacognitive tasks to AI systems, including goal setting, comprehension monitoring, and reflective evaluation, rather than engaging in these processes themselves. Researchers have termed this phenomenon metacognitive laziness.

A research team from The University of Hong Kong and The Education University of Hong Kong, alongside Monash University in Australia, has taken an initial step toward measuring this behavior. In a study published in of the journal ECNU Review of Education on May 15, 2026, the team reports the development and preliminary validation of the Metacognitive Laziness Scale (MLS), a six-item instrument adapted from an established work avoidance measure and recontextualized for AI-mediated learning environments.

“The introduction of GenAI into education represents a double-edged sword. While it offers unprecedented efficiencies in task completion, it may simultaneously undermine the metacognitive processes that are foundational to deep learning and academic achievement,” the authors note.

Items for the MLS were adapted from the Work Avoidance Goals subscale of the Goal Orientation and Learning Strategies Survey (Dowson & McInerney, 2004), with each item revised to capture AI-specific avoidance behavior. For example, the original item “I choose easy options in school so that I don’t have to work too hard” was reworded as “I choose to use AI for assignments, so I don’t have to think too hard.” Prior to data collection, two content experts independently reviewed and refined the item pool for relevance and clarity.
The scale was administered to 144 undergraduate health professions students across six disciplines at a Hong Kong university, following a three-week interprofessional education simulation course. The sample spanned nursing, pharmacy, medicine, Chinese medicine, social work, and food and nutritional science; it included students from all four undergraduate year levels. The authors acknowledge that the sample is modest in size and geographically limited, and that these factors constrain the generalizability of the current findings.

Initial psychometric analyses yielded promising but preliminary results. Confirmatory factor analysis supported a unidimensional factor structure, with all six items loading strongly onto a single latent factor and the scale demonstrating high internal consistency. A structural equation model examined the relationships between metacognitive laziness and students’ engagement and disaffection with learning. The MLS showed significant positive associations with both behavioral and emotional disaffection, accounting for a substantial portion of variance in these outcomes, while showing no meaningful relationship with engagement.

“We hope that our small effort to contribute to advancing our understanding of AI-driven metacognitive laziness will ignite traction among the community of practitioners to inform the enrichment of their research agenda,” the authors write.

The researchers are explicit about the preliminary nature of these findings. The cross-sectional design means causal inferences cannot be drawn, and it remains unclear whether AI use contributes to reduced metacognitive development over time or whether students with pre-existing tendencies toward effort avoidance are simply more inclined to use AI this way. The authors call for replication with larger, more diverse samples and measurement invariance testing across demographic groups.

The MLS is available for researchers and educators who wish to examine AI-mediated metacognitive offloading in their own contexts. As AI tools continue to reshape how students learn, instruments like the MLS offer a practical starting point for understanding when technology supports learning and when it may be getting in the way. The authors welcome the scale's uptake and adaptation across diverse educational settings and hope it will serve as a useful resource for those working to promote responsible and effective AI integration in education.


Reference
Dizon, J. I. W. T., Mendoza, N. B., Gasevic, D., et al. (2026). Assessing AI-Driven Metacognitive Offloading: Initial Development and Validation of the Metacognitive Laziness Scale. ECNU Review of Education, 9(2). https://doi.org/10.1177/20965311261450994

Funding information
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by the Teaching Development Grant from the University of Hong Kong, granted to the corresponding author.