An exploratory behavioral and electroencephalographic study of artificial intelligence-assisted learning modes in high school students

Synthesis: This study investigates how different modes of AI interaction affect cognitive engagement and learning outcomes in high school students. Using a within-subjects design with 48 students, the researchers compared three conditions: Auto mode (AI solves problems independently), Interactive mode (student-AI collaboration with Scaffolding), and Manual mode (no AI assistance). The Interactive mode produced the highest cognitive engagement and task accuracy, while the Auto mode led to reduced engagement and potential Over-Reliance.

The use of electroencephalography (EEG) provides a neurophysiological dimension to understanding AI's impact on learning. Though EEG results did not reach statistical significance, descriptive patterns suggested differences in neural activity across the three AI interaction modes. This connects to broader debates about student-AI interaction design in K-12 classrooms.

The finding that full automation reduces cognitive engagement echoes the Over-Reliance concerns documented in prior research on AI tutoring systems. The study's framework for categorizing AI interaction modes (Auto, Interactive, Manual) provides a replicable methodology for future human-AI interaction research in educational settings.

What this means for practice

• Learners. Choose the AI mode deliberately: behavioral profiles differed significantly across all three measures (Initiation, Processing, Stress) depending on whether the AI acted as tutor, collaborator, or solver, so the mode you pick changes how you engage.
• Learners. Ask for a tutor — steps and guidance only — rather than for the finished answer, and avoid solver mode, whose patterns the authors read as consistent with the offloading described in prior literature.
• Learners. Keep working problems yourself after using AI; the study explored whether AI use carries over to subsequent independent work and found only a descriptive pattern, so the effect of AI on your next task should not be assumed.
• Learners. Take a break between AI-supported and independent blocks: participants showed signs of mental fatigue (rising delta activity) as questions progressed, which the authors flag as a threat to their own measurements.

Limitations

• The study had 48 high school students (24 male, 24 female, ages 14–18) from a single large public high school in Texas, and all participants in the final volunteer sample self-identified as Asian.
• The sample was deliberately narrow: an academically uniform cohort with weighted GPA of 4.0+, at least one AP/IB class, bilingual from birth, and self-reported daily AI use for at least two years — so these are experienced, high-achieving AI users, not typical learners.
• The EEG data did not reach statistical significance (p > 0.29 for neural differences across modes) and the headset had limited spatial coverage, so the neural results are descriptive trends only.
• Fatigue accumulated across the quiz (descriptive increases in delta activity), and the authors recommend larger samples and separated task blocks to address participant-level variation in focus.

Citation

Kashika Khurana, Ally Liew (2026). An exploratory behavioral and electroencephalographic study of artificial intelligence-assisted learning modes in high school students. arXiv cs.HC / cs.CY.