<p>In this paper, we present a simulation of social network dynamics using agents powered by the GPT-4o large language model. These agents are designed with cognitive modules for memory, reflection, and chain of thought to better emulate human-like behavior. Focusing on the Roe v. Wade case, we model public opinion propagation on a simulated Twitter-like platform through actions like likes, retweets, comments, and posts. Multiple simulation experiments are conducted with 50 agents over 10 timesteps to observe their behavior. By employing small-world networks and scale-free networks as structural variables, we analyze the impact of different network topologies on information dissemination, interaction patterns, and the formation of echo chambers. Our comparative study reveals distinct patterns of opinion dynamics in each network type, highlighting the crucial role of network topology in shaping public discourse.</p>