The acknowledgment of occupant behaviour as a key driver of uncertainty in building energy analysis is today well established. Existing literature highlights the need of carefully addressing human-related interactions with the building envelope and systems. In response to this need, researchers have proposed a number of stochastic models that aim at reflecting occupant behaviour patterns in building energy simulation to bridge the gap between simulated and real energy consumptions in buildings. However, most proposed approaches for modelling occupant behaviour consider time-related factors and physical parameters such as indoor or outdoor environmental variables while less attention is paid to other influential factors such as psychological, social and contextual drivers or individual thermal comfort attitudes and preferences of the occupants. To understand occupant behaviour in a comprehensive manner, these factors should be carefully addressed in upcoming occupant behaviour models. The Bayesian Network framework presents a promising environment for hierarchically and flexibly structuring a large number of explanatory variables that drive the occupant to perform a certain action. This paper describes the development of a theoretical model of occupant's window control behaviour with an extensive set of drivers and highlights the capability and usability of Bayesian Networks to develop such models based on field measurements and information collected through surveys compiled by the building occupants.