In flood risk assessment, there remains a lack of analytical frameworks capturing the dynamics emerging from two-way feedbacks between physical and social processes, such as adaptation and levee effect. The former, “adaptation effect”, relates to the observation that the occurrence of more frequent flooding is often associated with decreasing vulnerability. The latter, “levee effect”, relates to the observation that the non-occurrence of frequent flooding (possibly caused by flood protection structures, e.g. levees) is often associated to increasing vulnerability. As current analytical frameworks do not capture these dynamics, projections of future flood risk are not realistic. In this paper, we develop a new approach whereby the mutual interactions and continuous feedbacks between floods and societies are explicitly accounted for. Moreover, we show an application of this approach by using a socio-hydrological model to simulate the behavior of two main prototypes of societies: green societies, which cope with flooding by resettling out of flood-prone areas; and technological societies, which deal with flooding also by building levees or dikes. This application shows that the proposed approach is able to capture and explain the aforementioned dynamics (i.e. adaptation and levee effect) and therefore contribute to a better understanding of changes in flood risk, within an iterative process of theory development and empirical research.