The Illusion of Cognitive Independence in the Age of Generative AI: Unpacking GenAI Literacy and Learner Profiles

Synthesis: In a survey of 848 academic learners recruited internationally, the researchers examined how GenAI literacy shapes the way students actually engage with AI tools. GenAI literacy—particularly critical-ethical awareness—worked through a dual pathway: it promoted critical monitoring of GenAI outputs while also enabling strategic reliance on GenAI as a cognitive shortcut. Cluster analysis distinguished three learner profiles whose patterns revealed an "illusion of cognitive independence," in which learners report strong independent thinking even while offloading core cognitive work to the machine.

Key Findings

• Three distinct learner profiles emerged. Cluster analysis grouped users into Engaged and Literate, Uncritical Reliant, and Skeptical Minimal. The first combined heavy GenAI use with strong operational and conceptual AI Literacy and moderate critical-ethical awareness; the second showed moderate use but low critical and conceptual literacy; the third used GenAI rarely yet relied on it heavily when they did engage.
• Critical-ethical awareness played a dual, counterintuitive role. It was the strongest predictor of critical monitoring of GenAI outputs, yet it also predicted greater shortcut reliance—that is, using GenAI to bypass independent cognitive effort. Greater awareness did not uniformly reduce reliance on the tool.
• The "illusion of cognitive independence." Uncritical Reliant users, who monitored least and scored lowest on literacy, nevertheless reported the highest preservation of independent thinking, while highly literate Engaged and Literate users reported the lowest. This mismatch between self-perceived autonomy and actual Cognitive Offloading persisted even after controlling for baseline academic self-efficacy.
• Skeptical Minimal users leaned on shortcuts. Despite barely using GenAI, they showed the highest shortcut reliance when they did engage, suggesting that low familiarity may produce unverified dependence rather than abstention.
• Literacy paired with monitoring predicted stronger outcomes. Engaged and Literate users reported the best problem solving, task execution, and cognitive endurance, consistent with GenAI functioning as a scaffold when matched with critical oversight.
• Conceptual literacy played a limited role in immediate behavior, while critical-ethical awareness proved far more behaviorally consequential—technical understanding mattered less than the inclination to evaluate and verify outputs.
• Gender differences were modest but systematic, with male learners reporting higher use frequency and operational/conceptual literacy and female learners reporting higher critical literacy and monitoring.

Study Design & Method

The study drew on an online questionnaire completed by 848 academic learners aged 18–50, recruited through the Prolific platform across dozens of countries after excluding respondents who reported no GenAI use. Participants rated their GenAI use frequency, their literacy across operational, conceptual, and critical-ethical dimensions, their usage behaviors (critical monitoring, general reliance, and shortcut reliance), and their self-perceived academic skills such as problem solving, task execution, cognitive endurance, and independent thinking. Hierarchical multiple regression examined how the literacy dimensions predicted immediate usage behaviors, while cluster analysis identified the learner profiles; follow-up comparisons then tested how the profiles differed on the skill outcomes.

What this means for practice

• Instructors. Teach critical monitoring and verification as explicit, assessed practices aimed at the two profiles most at risk: Uncritical Reliant learners, who monitor least, and Skeptical Minimal learners, who leaned on shortcuts whenever they did engage; for Engaged and Literate learners, target instead the sharper evaluative judgment about what is safe to delegate.
• Instructors. Make the gap between perceived and actual contribution visible through calibration activities, because Uncritical Reliant learners reported the highest preservation of independent thinking while offloading the most core cognitive work.
• Designers. Do not treat tool adoption or conceptual literacy as the outcome: critical-ethical awareness predicted more critical monitoring and more shortcut reliance at once, so set tasks with a deliberately non-delegable core and design for calibrated reliance — strategic delegation paired with epistemic oversight under Self-Regulated Learning — rather than assuming literacy reduces reliance.
• Researchers. Replicate the profiles with behavioral measures, since the illusion of cognitive independence is defined by the self-perception that the self-report instrument cannot independently verify.
• Researchers. Extend the profile comparisons beyond the graduate-heavy sample and the Southern and Eastern European recruitment base before treating the three profiles as general.

Limitations

• Self-reported measures of perceived skills and usage behavior mean the relationships may not reflect actual behavior or skill change—especially relevant given the illusion depends on self-perception.
• The cross-sectional design precludes causal claims about how literacy shapes behavior or skills over time.
• The instrument was largely novel and developed for this study, with one scale showing relatively low reliability.
• The sample skewed toward graduate students and toward Southern and Eastern Europe, limiting generalizability across educational levels and cultural contexts.

Connected Concepts

• AI Literacy — the competency profile the study decomposes into operational, conceptual, and critical-ethical dimensions.
• Cognitive Offloading — the delegation of mental effort that underlies the illusion of cognitive independence.
• Self-Regulated Learning — the framework positioning GenAI literacy as a regulatory mechanism of human-AI interaction.
• Higher Education — the setting and target population for the profile-based findings.
• Student Engagement — how patterns of GenAI use and monitoring reflect different modes of cognitive engagement.

Connected Articles

• Cognitive Offloading in the Age of Generative AI: What Does It Mean for Students With Learning Disabilities?
• Reclaiming Epistemic Agency: A Critical Framework for Human-Generative AI Co-Agency in Education
• AIGC Affordance and Student Self-Regulation: A Serial Mediation Model of Self-Efficacy and Learning Motivation
• Patterns of Learner-AI Interaction and Performance in an Object-Oriented Programming Course

Citation

Amzalag, M., Zviel-Girshin, R., & Beimel, D. (2026). The Illusion of Cognitive Independence in the Age of Generative AI: Unpacking GenAI Literacy and Learner Profiles. Frontiers in Education, 11, 1861548.