The extent of echo chambers in social media and their impact on polarization is an open debate in the political science literature. This paper employs an agent-based model to explain how echo chambers can emerge on social media platforms. Simulation conditions separately manipulated (1) the homophily or group similarity in personal networks and (2) the level of selective exposure through filter bubbles on social media. Individuals' information exposure significantly influenced echo chamber emergence, particularly where people are surrounded by a diverse group of others. Similarly, artificially creating echo chambers did not ensure their persistence over time in the presence of filtering algorithms. Overall, we need to reconsider how we understand echo chambers and the mechanisms responsible for this phenomenon.