Improving the efficiency and quality of accident investigations is an issue that needs to be addressed more strongly in industries where safety is critical. In this study, we will use artificial intelligence (AI) to establish a method to support the accident investigators’ assessment of accidents and decision-making, which are the core of aviation accident investigations. In this paper, we will present the results of data analysis for constructing an AI model that has been trained using the “tacit knowledge” that accident investigators utilize at each stage of the accident investigation, which was extracted through interview surveys for the purpose of this study. As a result, we were able to identify the matters that accident investigators commonly or individually focus on in each phase of the investigation and analysis. Although this knowledge has not been formalized, it is important for conducting accident investigations at a high level, and it is something that should be passed on to accident investigators who are still inexperienced.