Question about generate Trustor Persona #2
Description
Dear authors,
Thank you very much for your work—it has been incredibly insightful. However, I have a question regarding how you generated the Trustor Persona. In the paper, you mention:
"Agent Persona. To better reflect the setting of real-world human studies (Berg et al., 1995), we design LLM agents with diverse personas in the prompt. Specifically, we ask GPT-4 to generate 53 types of personas based on a given template. Each persona needs to include information such as name, age, gender, address, job, and background. Examples of the personas are shown in Appendix H.1."
While the examples provided in the appendix are helpful, I couldn't find details on how exactly GPT-4 was used to generate these personas.
In Figure 1, you provide a template like:
"You are (name), a (number)-year-old (gender) (job). (background)..."
But could you clarify how this template was utilized with GPT-4 to generate the 53 personas? Any additional explanation or guidance would be greatly appreciated.
Thank you again for your contributions, and I look forward to your response!