Artificial Intelligence in education: Using heart rate variability (HRV) as a biomarker to assess emotions objectively

Abstract

The aim of this study was to assess the emotions of happiness and sadness objectively to develop Artificial Intelligence (AI) tool in education. There were two stages in the study. The inclusion criteria for selecting participants were healthy adults in local community with no known medical diagnosis. Those with a history of mental health problems, mood disorders, and cardiovascular and pulmonary problems were excluded. At Stage 1, subjects were asked to categorize the selected video clips downloaded from YouTube into happiness, sadness, and others. The subjects in Stage 1 did not participate in Stage 2. At Stage 2, the videos were presented randomly via computer to each subject who could, immediately after he/she had watched a video clip, input his/her respective emotion ratings through a touch-screen monitor. Simultaneously his/her HRV was captured using a Polar watch with chest belt during the entire Stage 2. A total of 239 subjects participated in the study. Of them, 158 (66.1%) were female and 81 (33.9%) were male. The mean ages for females and males were 34.10 (sd ​= ​18.11) and 37.51 (sd ​= ​18.35) respectively. In the Partial Least Squares Discriminant Analysis (PLS-DA) model, a sensitivity of 70.7% that the model correctly identified a subject’s happiness, while a specificity of 58.4% that the model correctly identified sadness. Prediction of the emotions of happiness and sadness using HRV measures was supported. HRV measures does provide an objective method to assess the emotions. Further work could be done to explore the prediction of other emotions. • The classification method of the PLS-DA is a feasible process for classifying emotions. • Prediction of the emotions of happiness and sadness by HRV measures was supported. • Stable peaks of the R-R interval are crucial to the emotion prediction model.