Towards Modelling Self-imposed Filter Bubbles in Argumentative Dialogue Systems