EEG signals reveal engagement differences in preferred learning styles
Preference vs. Performance: EEG-Based Classification of Learner Engagement in Multimodal Instruction
Human-Computer InteractionComputers and Society
Summary
People often say they learn better when teaching matches their favorite style like seeing, hearing, or doing. The researchers found that while matching teaching to these preferences doesn't always improve test scores, it does change brain activity linked to paying attention. They used brainwave sensors to tell when someone was more engaged based on their preferred style. This shows brain signals can help computers figure out who is really focused during learning, even if it doesn't always mean better grades.
What this means in practice
- •For educational technology developers: Create adaptive learning platforms that monitor EEG signals to adjust teaching methods based on real-time engagement rather than self-reported preferences.$Commercial implications: Enables development of personalized learning systems that use affordable EEG devices to improve learner engagement tracking and adaptation.
- •For cognitive workload engineers: Use EEG engagement patterns to detect attention states during multimodal content delivery for optimizing human-computer interaction.
Authors
Sri Jahnavi Adusumilli, Deepak Giri, Pallavi Vaswani, Pallavi Singh, Megha Moncy, Lalitha Pranathi Pulavarthy, Saptarshi Purkayastha
Abstract
Effective adaptive instructional systems require robust measures of learner engagement that go beyond static user profiles. This study employs Multimodal Learning Analytics (MMLA) to investigate the relationship between self-reported instructional modality preferences and objective neurophysiological markers of engagement. Thirty-seven participants engaged with learning content delivered via varying modalities (visual, auditory, reading/writing, kinesthetic). We captured real-time neural activity using two EEG devices: the Emotiv EpocX (14 channels, 128 Hz) and OpenBCI (16 channels, 125 Hz). Preferences were assessed using the VARK questionnaire. Consistent with literature challenging the "meshing hypothesis," aligning instructional modality with stated preferences did not significantly predict performance gains. However, spectral analysis of EEG data revealed divergent engagement patterns: when content aligned with preferences, distinct neural activity patterns emerged in theta and alpha frequency bands-markers associated with attention and cognitive processing. These signals were used to train a binary logistic regression classifier, achieving a mean accuracy of 83.21% with OpenBCI and 56.27% with Emotiv EpocX. These findings suggest that while self-reported preferences may not dictate learning outcomes, they significantly influence neurophysiological engagement, offering a viable, non-invasive input for adaptive educational algorithms.