Presentation
A Comparative Analysis of Human and AI-Coded Helicopter EMS Accidents Using HFACS
DescriptionAccurate accident analysis is essential for aviation safety, yet coding human error using the Human Factors Analysis and Classification System (HFACS) remains labor-intensive and prone to inconsistency. This study investigates the potential of a customized AI model, HFACS GPT, to automate the classification of helicopter emergency medical services (HEMS) accident reports, comparing its performance to that of trained human analysts across 132 cases. Results suggest that the AI model demonstrates substantial alignment with human-coded data, particularly in identifying environmental hazards, skill-based errors, and supervisory issues. Notably, its performance showed measurable improvement as more examples were processed, indicating a positive learning curve and growing refinement in classification accuracy over time. While the AI occasionally struggled with complex distinctions, such as mental versus physical fatigue or nuanced decision errors, many of the mismatches were minor or within the same HFACS tier. Although full validation is ongoing, early findings support the model’s potential as a high-speed, reliable coding assistant. Rather than replacing expert judgment, the AI can serve as a first-pass analyzer, helping standardize and scale HFACS applications. This hybrid approach, combining machine efficiency with human expertise, could transform aviation accident analysis by improving consistency, reducing workload, and accelerating safety insights.
Event Type
Lecture
TimeTuesday, October 14th4:50pm - 5:10pm CDT
LocationGrand A
Safety
Similar Presentations