Mirroring the debate between abstraction and descriptiveness that has been going on in the philosophy of science for quite some time, the trade-off between the practicality and descriptive adequacy of social simulations is a complex and context-dependent one. Relying on the importance and practicality of simplicity in modelling following the “Occam’s Razor” argument, much of agent-based social simulation models end up being rather abstract and generic to be of use for public policy decision-making. We argue that injecting data - with the aim of providing realistic “what-if” scenarios for policy decision support - and incorporating a participatory research design involving communities and stake-holders - with the aim of establishing relevant behavioural and structural validity – may help impart necessary realism and relevance into agent-based social simulations for use in public policy-making. While a move towards data-driven descriptive models allows for more of the available evidence to be synthesized, stakeholder participation entails an organic process of development of the models in which the stakeholders both explicate and refine their understanding of the target systems and use the models to investigate alternative policy or other strategic options. Devised on the basis of data and developed by means of a process of empirical validation, the usefulness of data-driven agent-based social simulation models developed with stakeholder participation is that they support the development of a social process of policy and strategic analysis when forecasting and prediction is infeasible with respect to the relevant natural and social systems. Drawing from ongoing work on an agent –based model exploring the role of environment in generating socio-spatial health inequalities, we outline a framework for integrating data-driven and participatory approaches in a simulation research design, with a view to achieving relevant and realistic empirical embeddedness from a policy perspective.