A Multimodal Approach Combining Biometrics and Self-Report Instruments for Monitoring Stress in Programming: Methodological Insights

2025-07-02

AI Summary

This study addresses the susceptibility of conventional self-report instruments to subjective bias in stress assessment during programming tasks. We propose a multimodal stress monitoring paradigm integrating biometric sensing with psychometric scales to enhance objectivity and sensitivity. Specifically, electrodermal activity (EDA) sensors were employed to capture real-time physiological signals, complemented by pre- and post-task psychological questionnaires and structured interviews, enabling multi-source data fusion. Results revealed no statistically significant stress changes in self-reported measures, while interview feedback exhibited high heterogeneity; in contrast, EDA desynchronization peaks demonstrated statistically significant differences under time-pressure conditions, indicating superior sensitivity to implicit, transient stress responses. This work establishes a reproducible mixed-methods framework for stress measurement in software engineering and empirically validates the unique utility of physiological metrics—particularly EDA—in detecting subtle, momentary stress fluctuations that remain undetected by traditional subjective instruments.

Technology Category

• Cognitive Modeling & Cognitive Systems: Affective Computing
• Application Domains: Software Engineering
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Security and Privacy: Large-scale security measurements
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• User Modeling, Personalization and Recommendation: Metrics for user behavior and evaluating success

Abstract

The study of well-being, stress and other human factors has traditionally relied on self-report instruments to assess key variables. However, concerns about potential biases in these instruments, even when thoroughly validated and standardised, have driven growing interest in alternatives in combining these measures with more objective methods, such as physiological measures. We aimed to (i) compare psychometric stress measures and biometric indicators and (ii) identify stress-related patterns in biometric data during software engineering tasks. We conducted an experiment where participants completed a pre-survey, then programmed two tasks wearing biometric sensors, answered brief post-surveys for each, and finally went through a short exit interview. Our results showed diverse outcomes; we found no stress in the psychometric instruments. Participants in the interviews reported a mix of feeling no stress and experiencing time pressure. Finally, the biometrics showed a significant difference only in EDA phasic peaks. We conclude that our chosen way of inducing stress by imposing a stricter time limit was insufficient. We offer methodological insights for future studies working with stress, biometrics, and psychometric instruments.

Problem

• Compare psychometric stress measures with biometric indicators
• Identify stress-related patterns in biometric data during programming
• Assess effectiveness of time limits for inducing stress

Innovation

• Combines biometric sensors with self-reports
• Analyzes stress via EDA phasic peaks
• Tests time pressure as stressor