From the reconceptualisation of the safety event to an innovative systemic risk management approach

Abstract

Safety events occur spontaneously or unexpectedly without prior planning. The objective of this research was to demonstrate that safety events are conceptualised as dynamic combinations of co-occurring factors, both related and unrelated, within a specific space-time context. The empirical demonstration that a deviation from an intended process is a particular combination of planned and unplanned factors makes the safety event predictable, advancing the evolving field of safety thinking. Research has shown that Artificial Intelligence (AI) can potentially support the identification of these combinations. A data model solution grounded in the novel "safety event" definition enabled extracting interpretable patterns from merged and extensive datasets. This fostered an innovative and evolving risk management approach and strategy in predictive terms, spanning from hazard identification to risk monitoring, driven by insights derived from patterns. The data analysis of two aviation-based case studies explored safety issues within an ultra-safe system and generated clear and simple patterns. However, the predictive potential of the data model indicates that more extensive and diverse data sources could reveal more intricate patterns, better capturing the configuration of a safety event with their planned and unplanned factors. This potentially can change the game for risk management in practice, providing a further strategy in risk mitigation with the concept of 'breaking the risk pattern' (Baranzini et al., 2013; Baranzini & Zanin, 2015). The research also assessed the organisational impact of implementing AI-based solutions. A total of 11 semi-structured research interviews were conducted involving aviation professionals with different backgrounds and expertise. The Cube methodology (McDonald, 2017; McDonald et al., 2021) was deployed as part of a socio-technical systems analysis to deploy an AI-based solutions system change. The Thematic Analysis provided valuable insights into assessing risks at a specific time and determining the necessary steps to transition from the current state ('as-is') to the complexities of implementing AI-based applications in organisations ('to-be') to establish a set of prerequisites for an effective system change, supporting the identification of potential solutions. The research, in its complexity and multidisciplinarity, has demonstrated its ability to address the research questions, from the reconceptualisation of the safety event to the design of a data model and data analytics process for extracting interpretable patterns and how these patterns can provide information that can be applied to an innovative systemic approach in risk management. It highlights the inherent difficulties in the system for effective implementation of AI-based applications. The research concludes its work, opening the door to further new initiatives that can drive the change management process where organisations, people and technology are simultaneously involved.