Large Language Models (LLMs) exhibit the ability to adopt different personas and behaviors based on social context. This flexibility is achieved through the presence of personality subnetworks within the models, rather than relying solely on external knowledge or prompting methods.