Modelling driver decision-making at railway level crossings using the abstraction decomposition space

The objective of this paper is to cast users of railway level crossings as fexible and adaptive decision-makers, and to apply a cognitive systems engineering approach to discover new behaviour-based insights for improving safety. Collisions between trains and road vehicles at railway level crossings/grade crossings remain a global issue. It is still far from apparent why drivers undertake some of the behaviours that lead to collisions, and there remains considerable justifcation for continuing to explore this issue with novel methods and approaches. In this study, 220 level crossing encounters by 22 car drivers were subject to analysis. Concurrent verbal protocols provided by drivers as they drove an instrumented vehicle around a predefned route were subject to content analysis and mapped onto Rasmussen’s Abstraction Decomposition Space. Three key results emerged. First, when they realise they are in a crossing environment, drivers’ natural tendencies are to look for trains (even if not required), slow down (again, even if not required), and for their behaviour to be shaped by a wide variety of constraints and afordances (some, but not all, put there for that purpose by railway authorities). The second result is that expert decision-making in these situations does not describe a trajectory from high-level system purposes to low-level physical objects. Instead, drivers remain at intermediate and lower levels of system abstraction, with many loops and iterations. The fnal fnding is that current level crossing systems are inadvertently constraining some desirable behaviours, afording undesirable ones, and that unexpected system elements are driving behaviour in ways not previously considered. Railway level crossings need to be designed to reveal their functional purpose much more efectively than at present.