WELD: The First Naturalistic Long-Period Small-Team Workplace Emotion Dataset for Ubiquitous Affective Computing

A new research dataset, WELD, has been introduced, offering the first long-period, naturalistic workplace emotion data collected from a stable small team using fully passive sensing. This dataset, comprising over 730,000 facial expression probability vectors from 49 employees of a Chinese software company over 30 months, addresses a significant gap in affective computing research by providing unique longitudinal and relational insights into workplace emotions.

Why it matters

This development is strategically important as it provides an unprecedented resource for understanding long-term emotional dynamics within professional settings. Such data can inform strategies for optimizing workplace well-being, team collaboration, and productivity, leveraging insights derived from advanced affective computing. It also highlights the growing capability to passively monitor and analyze human emotional states over extended periods in real-world environments.

Key insights

• WELD is the first dataset to combine long duration (30.1 months), naturalistic workplace context, stable small-team social structure, and fully passive sensing protocols.
• The dataset includes 733,780 per-frame seven-class facial expression probability vectors.
• Data was collected from 49 employees of a Chinese software company between November 2021 and May 2024.
• It represents the longest naturalistic 'in-the-wild' emotion corpus and the only multi-year corpus supporting both within-individual longitudinal and within-team relational analyses on the same subjects.
• The collection method was designed to survive institutional review, indicating adherence to ethical data collection principles for research.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2510.15221