Unmasking Conversational Bias in AI Multiagent Systems