Occupational physicians should use artificial intelligence as a decision-support tool rather than relying on it solely to determine whether workers are fit for their jobs, according to bioethics experts. These decisions carry profound consequences for workers' employment, income, dignity, and privacy, making them too ethically complex to delegate entirely to algorithms.

What Makes Fitness-for-Work Decisions So Ethically Complicated?

Consider a crane operator arriving for a shift and being told an AI system has classified them as "unfit for work" based on medical notes, sleep data from wearables, and patterns of short absences. The decision might seem efficient from an administrative standpoint, but it carries serious consequences. A fitness-for-work determination can affect employment status, income stability, professional reputation, access to workplace accommodations, and even public safety.

The core problem is that fitness-for-work assessments are not simple yes-or-no questions. They require understanding the specific worker, their specific job, and their specific workplace at a particular moment in time. A person might be unfit for one role but perfectly capable of performing another, or they might be fit for their current job with modifications, restrictions, or assistive devices. An algorithm that outputs only "fit" or "unfit" cannot capture this nuance.

Where Can AI Actually Help Occupational Medicine?

AI is not inherently problematic in occupational health. When designed and deployed responsibly, AI systems can strengthen occupational medicine by helping physicians manage complex information more efficiently. These tools can assist with several important functions:

• Risk Prediction: AI can help identify patterns and flag potential health risks that might otherwise be missed in large datasets.
• Data Organization: AI systems can organize and synthesize complex medical information, wearable data, and workplace hazard information into coherent summaries.
• Occupational Surveillance: AI can support ongoing health surveillance, injury severity estimation, and workplace hazard detection.
• Monitoring and Protection: AI can power wearable monitoring systems and smart personal protective equipment that track worker safety in real time.

The key distinction is that AI should remain a tool that supports human judgment, not replace it. When occupational physicians use AI responsibly, it can improve speed, consistency, and preventive action, especially when health teams must evaluate large amounts of data under time pressure.

Why Human Judgment Cannot Be Fully Automated?

Several ethical and practical reasons make it impossible for AI alone to make fitness-for-work decisions. First, AI systems rely on available data, but available data are not always morally relevant. Important information may be missing, outdated, inaccurate, or biased. Factors like supportive supervision, recent symptom improvement, a modified task plan, or the feasibility of a gradual return to work may not be captured in datasets but could be crucial to a fair assessment.

Second, respect for worker autonomy demands meaningful human engagement. Workers should understand how information about them is being used, have the opportunity to correct inaccurate or incomplete data, and be able to discuss their capacity and concerns with a qualified professional. Sole reliance on AI weakens autonomy because workers may not know what data was used, how the algorithm reached its conclusion, or how to challenge the result. This concern is especially acute in occupational medicine, where workers may feel pressured to accept health monitoring or algorithmic evaluation when refusal could threaten their employment.

Third, confidentiality and privacy are at stake. Occupational health data are sensitive because they can influence employment opportunities, workplace relationships, stigma, insurance decisions, and managerial actions. AI can intensify privacy risks by combining medical records, wearable data, absenteeism patterns, productivity indicators, and location data into predictions that appear objective. This creates the danger of "function creep," where data collected for health protection may later be used for productivity monitoring, workforce selection, or exclusion of workers perceived as costly or risky. Confidentiality obligations cannot be delegated to an algorithm or vendor; the occupational physician must remain responsible for deciding what information is relevant and ethically permissible to disclose.

How Should Occupational Physicians Integrate AI Responsibly?

The question is not whether occupational physicians should use AI at all, but how they should use it ethically. Responsible AI integration requires several safeguards:

• Transparency and Validation: AI systems must be transparent about how they reach conclusions and validated against real-world outcomes before deployment in high-stakes decisions.
• Clinical Supervision: AI tools must be integrated into professional judgment, with a qualified occupational physician reviewing and taking responsibility for all fitness-for-work determinations.
• Contextual Assessment: Physicians must understand the worker's specific tasks, physical and psychological job demands, workplace hazards, shift patterns, and available accommodations to interpret AI outputs in context.
• Error Recognition: Occupational physicians must recognize that AI errors are ethically significant because model outputs can be affected by poor-quality data, model drift, lack of external validation, or deployment outside the population for which the system was developed.

Erroneous fitness-for-work decisions can harm both workers and others. A false "fit" decision may expose a worker, coworkers, patients, or the public to preventable harm. A false "unfit" decision may unjustly exclude a worker from employment, worsen financial insecurity, create stigma, or delay recovery and return to work.

The stakes in fitness-for-work decisions are simply too high to treat them as purely technical outputs. These assessments sit at the intersection of health, livelihood, fairness, and public safety. They require individualized assessment, knowledge of actual job demands, consideration of reasonable accommodations, protection of medical confidentiality, and professional accountability. AI can be a valuable ally in occupational medicine, but it cannot replace the human judgment that these complex, consequential decisions demand.