In this talk we will give an introduction to Agent-Based Modelling (ABM) using Python. ABM is a technique used to build simulated populations of heterogeneous individuals that are governed by rules of behaviour. These individuals, or agents, interact in simulated environments in which we study their emergent behaviour. We build agent-based models in order to try explain why people do what they do. In this talk, we'll briefly show how we develop and validate agent populations. We will then present on-going work modelling the 2011 London riots, investigating how riots spread and in particular focussing on police response. Finally, we’ll discuss why we use Python and the PyData stack to develop models, and mention how we are starting to use machine learning (scikit-learn) for learning rules of behaviour.