Estimation for cognitive load in Video-based learning through Physiological Data and Subjective Measurement by Video Annotation

When designing a video-based learning such as MOOC, it is very important to understand the cognitive aspects of learning and reflect them in the design. Many studies use subjective and physiological data as indicators of cognitive load. To fully understand the cognitive load, we need to understand both of them simultaneously. Therefore, this study is to investigate whether eye data(Mean Pupil Dilation, Mean Fixation Duration) predicts subjective cognitive load during video learning. Furthermore, as a second research question on a broader scale, we examined whether eye data predicts high and low states of subjective cognitive load during video learning. Through this, we expected to find the possibility of Video Annotation and Eye data as a way to measure Cognitive Load during video learning. The experiment was conducted in a controlled laboratory environment with 100 students. In the video learning situation, the learner's eye data was measured using an eye tracker. Immediately afterwards, a video annotation(VA) interview technique was used to put markers according to the cognitive load types such as A(Understanding), B(Easy), C(Complicated), and D(Discomfort). The collected data will be analyzed by Support Vector Machine, a machine learning technique that is considered appropriate for the physiological data.