Leveraging LLM-based agents for social science research: insights from citation network simulations