In this paper, we develop an agent-based model to explore how agents’ activity patterns affect echo chamber formation when their mutual interactions are controlled using a personalization system algorithm that decides what information users will be exposed to. In our model, agents can undertake two types of actions: publish a post and like a post. Our experiments revealed that the key parameter that guides agents’ opinion dynamics is the probability of publishing a post: agents who often publish posts tend to enter echo chambers. In contrast, the roles of network topology and liking behavior are far less influential.