Participatory inverse modelling for behavioural insights into Dutch EV charging dynamics

This paper presents a general epistemic workflow for behavioural inference in agent-based models under structural-parametric uncertainty by combining participatory and inverse modelling. This replicable computational pipeline enhances model realism, results explainability, and interpretability by integrating domain expertise. We apply our method to study Dutch EV charging behaviour. Using behavioural domain expertise, we co-design experiments and interpret alternative behavioural dynamics evaluated through inverse modelling in a structural-parametric pipeline. Public EV charging demand over a week in Den Haag is matched by simulating demand in an existing model of EV charging behaviour. Our method reveals that EV charging is largely driven by range anxiety or a lack of consideration for the availability of excess energy in the grid. While our behavioural findings are case-specific, the workflow is applicable to any agent-based model with ambiguous behavioural mechanisms and suitable for feeding back into theory, scrutinising formalisation, identifying areas for model improvement, and raising considerations for policy.