Physiology vs. Perception paradox: exploring GSR-based and questionnaire stress responses

• Main Authors: Md Nasir, Muhammad Syamil Tamlikha, Zulkiffli, Kamelia Zahiah, Ghazali, Aimi Shazwani, Jasni, Farahiyah
• Format: Proceeding Paper
• Language: English; English
• Published: IEEE 2024
• Subjects: T Technology (General); TA329 Engineering mathematics. Engineering analysis
• Online Access: http://irep.iium.edu.my/115383/1/115383_Physiology%20vs.%20Perception%20paradox.pdf; http://irep.iium.edu.my/115383/7/115383_Physiology%20vs.%20Perception%20paradox_SCOPUS.pdf; http://irep.iium.edu.my/115383/; https://ieeexplore.ieee.org/document/10652491

Stress, a prevalent mental health concern, is a complex phenomenon to quantify. This study explores the impact of stress induction method variations on Galvanic Skin Response (GSR), an electrophysiological measure, in differentiating relaxation and stress states. We employed the Stroop Color Word Test, Digit Span Test, and Mental Arithmetic Test as established stress induction methods, also known as stressors. To explore stress level variations, modified versions of the stressors, incorporating unpleasant visuals and loud audio were also developed. A mixed-methods design with twelve participants utilized GSR alongside questionnaires (subjective measures): the State-Trait Anxiety Inventory and the Psychological Reactance Test to assess subjective stress. Our findings revealed significant differences between relaxation and stressful conditions, confirming the validity of GSR for stress detection using the ANOVA test, F(6, 60) = 2.563, p < 0.05. Interestingly, a significant discrepancy was revealed between the physiological stress response (measured by GSR) and participants' reported stress levels. The GSR data indicated the highest stress response for the Digit Span Test with the mean of 0.415 (for original) and the mean of 0.411 (for modified), although participants reported finding the modified Mental Arithmetic Test to be more stressful via questionnaire (mean = 2.4417). This highlights the potential limitations of relying solely on subjective measures. Future research will involve more participants and explore integrating machine learning for a more robust system. This could lead to a comprehensive stress assessment tool with personalized interventions.