ARTIFICIAL INTELLIGENCE IN OCCUPATIONAL SAFETY AND HEALTH: TRANSITION FROM REACTIVE MODEL TO PREDICTIVE RISK MANAGEMENT

Abstract

This article analyzes the paradigm shift in Occupational Safety and Health (OSH) from traditional reactive approaches, focused on post-factum incident analysis, towards proactive risk management models powered by Artificial Intelligence (AI) and Big Data. The reactive OSH model, based on audits and lagging indicators, is critiqued as labor-intensive, subjective, and incapable of incident prevention. The need for AI is driven by the complexity of modern challenges, including psychosocial factors, remote work issues, and complicated ergonomic risks, which demand dynamic, real-time data processing.The transformation requires the OSH specialist to move from the function of an 'inspector' to a 'data-driven risk manager'. AI, encompassing methods from video analysis to text processing, automates cognitive functions, while Machine Learning (ML) forms the basis of predictive analytics. ML models, such as Classification Trees, Support Vector Machines (SVM), and Artificial Neural Networks (ANN), significantly surpass traditional statistical methods by identifying complex, non-linear dependenciesbetween work conditions and incident probability.The technical architecture relies heavily on two areas:1. Computer Vision (CV): Utilizing Deep Learning models like YOLO (You Only Look Once), CV systems analyze video streams in real-time. Key applications include monitoring Personal Protective Equipment (PPE) compliance, achieving up to 95% adherence on thermal power plants, and detecting unsafe behavior (e.g., falls, unauthorized access).2. Natural Language Processing (NLP): Algorithms like Topic Modelling and the Naive Bayes classifier process unstructured textual reports (injury narratives, near-misses) to automatically extract typical patterns and root causes that are often missed in manual processing. Crucially, AI facilitates the automated detection of Near-Misses (leading indicators). Automated registration generates a vast, objective data set, overcoming subjective barriers associated with manual documentation.Empirical evidence from the chemical industry demonstrates an 800% increase in near-miss reporting following AI  implementation, proving the operational success of the proactive transition. This integrated data  collection allows for a "closed-loop" system for dynamic safety management.However, the article emphasizes considering the psychosocial consequences of continuous monitoring. Based on the Stress Proliferation Theory, a link is identified between constant digital surveillance and increased employee anxiety.Practical recommendations for enterprises include prioritizing data digitization and immediate NLP implementation, initiating pilot CV projects in critical areas, and establishing clear ethical policies founded on transparency and maintaining human control. By adopting these principles, OSH systems can transition from reactive analysis to continuous, predictive risk mitigation.