Perceived AI agency is associated with self-reported reflection

Updated Sep 07, 2026

student voicefeedback

Aniekan Essien, Xue Zhou, Marios Kremantzis and Da Teng examine perceived control when students use AI for learning. Their 2026 study distinguishes reported agency from a direct measurement of learning.

What the survey found

The analysis used responses from 145 students in UK-based and 164 in China-based higher education. The groups describe study settings, not homogeneous nationalities or nationally representative populations.

Students reporting stronger initiative, monitoring and decision-making also reported more reflection; reflection was associated with self-reported critical thinking. The reflection–academic-self-concept relationship was significant in the UK sample but not the China sample. AI literacy was examined as a moderator in the China sample.

The authors explicitly reject causal interpretation. Contextual explanations were not directly measured, and only partial measurement invariance was established, limiting comparisons. No behavioural assessment demonstrates improved thinking or reduced dependence on AI.

Questions for institutional feedback

Our recommendation is to ask separately about checking outputs, making final decisions and perceived usefulness. Invite examples of what students accepted, rejected or changed. Treat those descriptions as reported experience, then choose additional evidence if the decision concerns learning outcomes.

Governed comment analysis can organise themes for discussion. It cannot determine a student’s cognitive dependence or diagnose why an outcome occurred from text alone.

FAQ

Does staying in control of AI cause better critical thinking?

This study does not establish that. Its model concerns associations among self-reported constructs.

Can the two samples establish cultural differences?

No. Study setting combines several possible influences, while cultural mechanisms were not directly tested.

Reference

Aniekan Essien, Xue Zhou, Marios Kremantzis and Da Teng (2026). The agency gap: perceived human AI agency, reflection and generative AI learning across UK and China based higher education contexts. Studies in Higher Education.

Correction, 7 September 2026: replaced the causal headline, clarified study location and self-report limits, and added the measurement-comparison caution. Verification covers accessible indexed sections, not the complete paper.