A Multimodal Dataset for Assessing Emotion, Stress, and Emotional Workload in Interpersonal Work Scenario

Area of Science:

• Affective Computing
• Human-Computer Interaction
• Psychology

Background:

• Workplace interactions often involve emotional labor, where employees manage expressed emotions.
• Existing datasets lack the complexity of real-world emotional dynamics in interpersonal work settings.
• Affective computing systems struggle with discrepancies between internal feelings and external expressions.

Purpose of the Study:

• To introduce EmoWork, a novel multimodal dataset for emotion and stress detection.
• To provide a resource for developing affective computing systems that understand nuanced workplace emotions.
• To address the gap in datasets capturing affect in naturalistic interpersonal work scenarios.

Main Methods:

• Collected multimodal data (physiological, behavioral, self-reported) from call center workers.
• Simulated realistic customer service interactions using professional actors.
• Utilized wearable sensors (Polar H10, Empatica E4, Muse S) and video/audio recordings.
• Extracted features for classification based on self-reported affective ratings.

Main Results:

• Developed EmoWork, a comprehensive dataset of emotion and stress in work settings.
• Included synchronized physiological, behavioral, and self-reported data.
• Captured data from simulated customer service interactions with dissatisfied customers.

Conclusions:

• EmoWork offers rich, context-specific data for advancing emotion and stress detection.
• The dataset enables research into affective computing for realistic workplace challenges.
• Facilitates the development of systems that account for emotional regulation in service occupations.