The article outlines the process of creating synthetic user research using Autogen, an autonomous agent orchestration tool. The application of the Large Language Model (LLM) from OpenAI was explored with versions GPT-3.5 and GPT-4. The whole process starts with setting up the environment and creating the Autogen configuration, LLM, and API keys.
The LLM instance has to be configured for tying each agent, which gives the flexibility to generate unique LLM configurations per agent. The article also discusses defining a researcher persona or agent for facilitating the simulated user research scenario, mentioning important system prompts, its role, and how the simulation ends.
Several customer personas are defined to be placed into the research panel. The article offers an example of system prompts generated with the help of generate_notice() function for staying on task.
The next steps involve defining the simulated environment and rules for speaker selection, adding the agents to group chat, and assigning a manager to manage the simulation. Human interaction is set, which allows passing instructions to various agents.
After the simulation, actionable insights are fetched which help in creating a summary and a Q&A scenario. The article then talks about gathering all conversations from the simulated panel discussion for using it as user prompt for the summary agent. A structured summary of the key findings such as pain points, preferences, and suggestions for improvement are generated.
Finally, a summary agent is defined along with its environment for running the summary. The article concludes with the output which appears in the form of a report card in Markdown and the ability to ask further questions in a Q&A style chat-bot. This experiment provides synthetic user research to gather insights about customers’ needs and suggestions based upon their distinct personas.
