The Personality Trap: How LLMs Embed Bias When Generating Human-Like Personas | ResearchPod