Interpersonal communication is supported by many new technologies nowadays. It has important influence on information diffusion. To study the dynamics of interpersonal communication, an agent-based model is built. The model simulates the process that agents construct their global cognitive social network through communication with local neighbors. Agents communicate with each other through social links and rewire their links to get closer to the information source occupying a better position in the network. So the dynamic structure of social links is a result of communication. Structure features based on different communication rules are compared. Positive feedback between agents' position and communication behavior is studied. Close neighbors' similarity is detected from outputs of the simulation. The research provides us better understanding about dynamic interpersonal communication process.