Agent-based simulation is an alternative approach to traditional analytical methods for understanding and capturing dierent types of complex, dynamic interactive processes. However, the application of these models is currently not common in the eld of socio-economical science and many researchers still consider them as intransparent, unreliable and unsuitable for prediction. One of the main reasons is that these models are often built on architectures derived from computational concepts, and hence do not speak to the selected domain's ontologies. Using Triandis' Theory of Interpersonal Behaviour, we are developing a new agent architecture for choice model simulation that capable of combining a diverse number of determinants in human decision-making and being enhanced by empirical data. It also aims to promote communication between technical scientists and other disciplines in a collaborative environment. This paper illustrates an overview of this architecture and its implementation in creating an agent population for the simulation of mobility demand in Switzerland.