Emergence and non-emergence for system safety

Abstract Emergence has been proposed as an important construct for research on the safety of sociotechnical systems. There is, however, some dissension about the fundamental nature of emergence. Furthermore, there is little clarity on how this construct might be used within safety research to guide analysis or design of sociotechnical systems. Emergence has proven to be a challenging construct to pin down in disciplines such as philosophy and computer science. Most troubling is that the distinction between emergent and non-emergent phenomena remains unclear. Here we offer a pragmatic view by outlining models of different types of emergence. We subsequently argue that one type, functional–semantic emergence, bears on a crucial distinction between Safety I and Safety II as discussed within system safety research. We conclude that safety of sociotechnical systems can be enhanced by integrating retroactive control of non-emergent phenomena with proactive control of emergent phenomena where retroactive control is achieved through use of rules and procedures and proactive control is achieved through attention to subtle information and use of recognition-primed decisions. Practitioner Summary Emergence has recently been invoked as an important construct in the systems safety literature. There is, nevertheless, some dissension in that literature regarding the nature of emergence and there is little explanation of how it can inform an approach to system safety. Here we contrast different models of emergence as a means of clarifying the nature of emergence and non-emergence. From there, we argue that the contrast between emergence and non-emergence can inform the contrast between Safety I and Safety II approaches to system safety.