Air Accident Analysis with AI: Reassessing Flight BA 5390 Using the Accimap and STAMP/CAST Methodologies

This article explores the potential of artificial intelligence (AI) in aiding the investigation of aviation accidents, particularly through the application of the Accimap and STAMP/CAST methodologies. These frameworks are recognized for their ability to map complex causal relationships in aviation incidents, providing insights into the structural, mechanical, organizational, and human factors involved. The study focuses on the BA 5390 accident, utilizing AI to generate textual and schematic representations of the events and contributing factors. The AI was guided through structured prompts to perform tasks such as hazard identification, control structure modeling, and the generation of safety recommendations. While AI demonstrated strengths in text-based analysis and logical structuring, challenges emerged in generating accurate graphical representations, especially in the Accimap methodology. Human intervention was required to validate and refine both the textual outputs and the diagrams, ensuring factual accuracy and addressing gaps in the AI's understanding of complex system interactions. The findings suggest that AI can be a valuable tool in accident investigations, offering efficiency in organizing and processing large datasets, but it requires human oversight to mitigate potential inaccuracies and to deepen the analysis of human factors. The study concludes that AI is most effective when used in conjunction with expert judgment, particularly in scenarios where human factors and decision-making processes are critical to understanding the full scope of an accident.

Keywords: Accimap, STAMP/CAST, Investigation, Artificial intelligence, Accidents, Aviation.