In this talk, we will analyse two models of opinion and behaviour changes regarding their tipping behaviour, that is, regarding drastic changes in the dynamics. Since agent-based models are usually very high-dimensional, we start by reducing the size of the state space. We search for collective variables that allow us to describe only the main degrees of freedom of the dynamics. Finally, we can study the tipping pathways in the reduced state space by means of Transition Path Theory, a method for computing statistical properties of the tipping paths such as their distribution, duration and rate.