abstract
- The main goal of this research is to propose a quantitative approach to identify critical thinking in undergraduate engineering students by leveraging electroencephalogram (EEG) measurements, facial emotion recognition and machine learning techniques. Measurements were taken by recording the face of the subjects and acquiring EEG signals from a wearable device. We explore the use of chaotic descriptors such as Lyapunov exponents, fractal dimension, Hurst exponent and approximate entropy to extract meaningful information from EEG signals. Subsequently we tested different machine learning algorithms to classify EEG signals and analyze the onset of critical thinking. An experiment was conducted in which students engaged in problem-based learning scenarios while wearing portable EEG devices. The collected data were then analyzed to discern patterns and discriminate between relevant, irrelevant, and false information in EEG signals. The experimental setup involved measuring the EEG signals from engineering students as they solved problem-based learning scenarios. The results demonstrated the ability to distinguish between different sections of a video based on brain activity, with false, relevant, and irrelevant information. Some findings include the left frontal electrode with the theta frequency band exhibiting the highest distinguishability for relevant information. From the experiments, we identified the Hurst exponent and approximated entropy as the most relevant descriptors for classification. The main contribution of this study is to offer an approach to decode critical thinking in engineering undergraduate students using EEG measurements and machine learning techniques. © 2024 IEEE.