This study investigates whether Electroencephalogram (EEG) biofeedback using consumer-grade wearable EEG headsets can serve as an effective learning analytics tool, providing feedback to students and teachers on cognitive and affective states during specific learning activities. The ultimate goal is to improve these cognitive states for better engagement and learning outcomes.

The emergence of wearable consumer-grade sensors, such as smartwatches, smart rings, and EEG headbands, has made it feasible to integrate these technologies into educational settings. Unlike medical-grade EEG devices, consumer-grade headsets are affordable and practical but may offer lower signal resolution and quality. This raises key questions: How accurately can consumer-grade EEG devices recognize complex cognitive and emotional states? And for which mental states are they most effective?

Previous studies have demonstrated success in predicting simple cognitive activities, such as relaxation or reaction to pain, but struggle with more complex tasks like reading or writing due to factors such as individual variability in EEG signals and the computational complexity of the data. This study leverages modern machine learning algorithms to determine whether group-based EEG analysis—treating multiple participants as one “collective brain”—can enhance predictive accuracy compared to traditional individualistic approaches.

Data was collected from eight participants using the Muse EEG headband during two sessions of classroom activities: watching a lecture, reflective writing, reading, and a quiz assessment. These sessions, conducted on two different days with shuffled activity sequences and new materials, minimized session-specific variations. During each activity, five minutes of EEG data was collected simultaneously from all participants. Topics were carefully selected to be engaging and relatively novel to ensure active participation.

Session 1 included a lecture on topic 1, reflective writing on topic 1, reading on topic 2, and a quiz on topic 2. Session 2 included reading topic 3, reflective writing on topic 3, a lecture on topic 4, and a quiz on topic 4. Activities were categorized as active (e.g., reflective writing, quiz) or passive (e.g., reading, lecture), with passive activities followed by active ones to encourage reflection and application.

After data cleaning and analysis, various machine-learning models were used for classification. Results showed high prediction accuracy: lecture watching (83.3%), reading (100%), reflective writing (94.7%), and quiz assessment (90.9%). These findings demonstrate that consumer-grade EEG devices can effectively predict cognitive activities in classroom settings. The group-based approach proved particularly effective, supporting real-time cognitive feedback for adaptive teaching and enhanced engagement.
Future research will expand classroom applications of EEG-based feedback, focusing on predicting cognitive states such as engagement, confusion, boredom, problem-solving, and mind-wandering to further optimize learning experiences.

Keywords: Education, education technology, learning analytics, EEG, wearable technology.