Impact of AI-driven Safety Systems on Worker Well-being
How to cite
Nikolova, Milena; Milanova, Viliana. 2025. "Impact of AI-driven Safety Systems on Worker Well-being." AEA RCT Registry. AEARCTR-0016911. https://doi.org/10.1257/rct.16911-1.0
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
This project investigates how people perceive workplace safety, well-being, and fairness when safety systems are managed by humans, artificial intelligence (AI), or a combination of both. The study is designed as a survey experiment with vignettes, fielded in October-November 2025 in the Netherlands via the probability-based LISS panel. A total of 2,100 adult respondents are pre-randomized into three groups: about 700 read a vignette describing a factory with an AI-only safety system, 700 read about a Human-only safety system, and 700 read about a Hybrid AI+Human system. Randomization is preloaded by Centerdata to ensure balance across age, gender, education, income, employment status, and urban/rural location.
The project is organized into two studies. Study 1 focuses on safety-related perceptions. Outcomes include how safe or risky the job is, whether it is stressful or physically demanding, and the likelihood of accidents, physical health problems, or mental health problems. These outcomes are measured before and after the safety system vignette, enabling within-subject comparisons. In addition, after the vignette, we measure between-subject outcomes: trust in the system, perceived effectiveness, and respect for workers’ privacy and dignity. As a placebo, we assess whether views on the societal importance of the job change with the introduction of the safety system.
Study 2 examines job quality. Outcomes include satisfaction with the job, its meaningfulness, and perceived fair wages. These are elicited before and after the vignette, allowing within-subject analysis of changes across the three treatment arms.
The survey takes about nine minutes to complete and was programmed and tested by Centerdata, which runs the LISS panel. Respondents are drawn from a representative sample of the Dutch population, ensuring comparability of treatment groups and reliable generalizability of findings.
The study has been funded by the Lloyds Register Foundation.
Intervention
Respondents take part in a survey experiment with vignettes. Everyone first reads a short description of a factory worker and standard workplace safety rules. At this point, after the baseline vignette, the respondents assess the baseline working conditions in terms of safety and job quality. They are then randomly assigned to one of three groups (about 700 respondents per group):
AI-only safety system: The factory introduces a system based entirely on artificial intelligence.
Human-only safety system: The factory introduces a system based entirely on trained human safety supervisors.
Hybrid AI + Human system: The factory introduces a system that combines both approaches, with AI monitoring and human supervisors.
After reading the vignette, respondents answer questions about safety, risks, trust, health, privacy, and job quality. Trust and privacy are only asked after the treatment.
Experimental design
This study is a survey experiment with vignettes conducted in the Dutch LISS panel. A total of 2,100 adult respondents are randomly assigned in advance to one of three groups of roughly equal size: (1) an AI-only safety system, (2) a Human-only safety system, or (3) a Hybrid AI + Human system. All respondents first read a short description of a factory worker under standard conditions. They then read one of the three safety system vignettes depending on their assigned group.
Some outcomes (safety, risk, health expectations, job satisfaction, meaningfulness, and fair wages) are measured both before and after the vignette, allowing for within-subject comparisons of changes. Other outcomes (trust in the system, perceived effectiveness, and respect for privacy and dignity) are measured once, after the vignette, and are compared between groups. Randomization is implemented by Centerdata to ensure balance across demographic characteristics.
Design details
Cited by
No citation data available for this registration yet.