@article{CRSIJ26000206,
                  author = {Emmanuel Ijeoma},
                  title = {Building AI-Assisted Risk Scoring Models for Proactive Safety Management},
                  journal = {Cosmo Research and Science International Journal},
                  year = {2025},
                  volume = {1},
                  number = {6},
                  pages = {496-502},
                  issn = {3108-1584},
                  url = {https://cosmorsij.com/published/CRSIJ26000206.pdf},
                  abstract = {For high risk sectors, the typical high impact incident is not caused by a single catastrophic event. More frequently, they are the result of "noise" in the signals that exist in everyday work, but are not strong enough or organized enough to be detected. Traditional safety systems focus on reporting incidents, audit, inspection and lagging indicators to identify an event once it has happened. These methods, while effective in enhancing organisation learning, are constrained by human recognition, inconsistent data interpretation and post hoc analysis. This type of situation introduces a critical time lag between the risk itself and when it becomes visible in the business. This paper discusses evolution and implementation of Artificial Intelligence (AI) aided risk scoring systems for proactive hazard identification and safety management. Utilising recent developments in machine learning, NLP, computer vision, and predictive analytics, the study investigates the possibilities of operational data sources such as near-miss narratives, Permit-to-Work records, environmental sensor data, inspection logs and behavioural indicators being combined to enable the identification of early patterns of increased risk conditions before incidents occur. The research brings evidence from predictive safety research literature as well as operational knowledge from the oil and gas and pipeline building and manufacturing sectors and considers both the technological and organizational aspects of predictive hazard intelligence. The results show that AI-driven systems have a marked ability to catch very subtle and non-obvious risk patterns that are not noticed through regular safety management routines. The usefulness of predictive systems, however, is demonstrated to be not only dependent on the accuracy of the algorithms, but also on the interpretability, contextual validation, trust, and human supervision. The study shows that the most value predictive intelligence has when being implemented as part of the operational decision making process, for example, Permit-to-Work systems, Go/No-Go assessments and frontline safety reviews.},
                  keywords = {Artificial Intelligence in Safety, Predictive Hazard Intelligence, Risk Scoring Models, Safety Systems Integration, Machine Learning, Human-AI Interaction, High-Operational-Risk Industries, Proactive Safety, Operational Risk Prediction},
                  month = {May}
        }