Stress at work hurts people and performance, but it often goes unnoticed until it is serious. This study explores whether deep learning can help organisations identify stress early and balance workloads more fairly.
The problem
Human resource management tries to get the best from people by balancing workloads, and managing stress is central to that. Yet AI, which has transformed many sectors, has barely been used in this area.
What the research did
We designed a deep neural network that classifies employees as stressed or satisfied using quantitative workplace measures. The method is designed to be general and adaptable, and to rely on measurable data to reduce bias in classification.
What we found
The network classified stressed and satisfied employees with 88.40% accuracy and an F1-score of 0.8728, outperforming similar approaches.
Who benefits
- HR teams wanting early signals to rebalance workloads
- Employees, through earlier support
- Organisations aiming for healthier, more productive workplaces
Limits and what comes next
Using AI to assess employees raises serious questions about privacy, consent and how results are used. Any real use should be transparent, voluntary and aimed at support, never punishment. The study is a first step, not a ready product.
The team
This paper was led by Nikhil Patel (University of Dubuque) with Sandeep Trivedi (Deloitte) and me.
Frequently asked questions
Can AI detect stress at work?
In this study, AI classified stressed and satisfied employees with 88.4% accuracy from quantitative measures.
Is it ethical to use AI on employees?
Only with transparency, consent and a clear focus on supporting staff. Privacy safeguards are essential.
What could organisations do with the results?
Spot overload early and rebalance workloads before stress becomes serious.