What the heck is it doing? Better understanding human-machine conflicts through models

This paper deals with human-machine conflicts with a special focus on conflicts caused by an “automation surprise”. Considering both the human operator and the machine autopilot or decision functions as agents, we propose Petri net based models of two real cases and we show how modelling each agent’s possible actions is likely to highlight conflict states as deadlocks in the Petri net. A general conflict model is then be proposed and paves the way for further on-line human-machine conflict forecast and detection