Synthetic Users

How does Synthetic Users work?

Synthetic Users uses the power of LLMs to generate users that have very high Synthetic Organic Parity. We start by generating a personality profile for each user, very much like a reptilian brain around which we reconstruct its personality.‍Once the personality profiles are established, these Synthetic Users are set in motion to interact in a simulated environment. Leveraging multi-agent frameworks, these users engage in conversations, make decisions, and evolve over time based on their interactions.

What are the data sources used by Synthetic Users?

Synthetic Users utilizes GPT, LLaMA, and Mistral foundation models. Users can also upload their proprietary data via RAG to enhance the specificity of the synthetic users and studies.

How is Synthetic Users different from ChatGPT?

Synthetic Users relies on an agentic architecture where multiple agents interact to fulfill a task. This, combined with the use of more than one model, adds diversity. Also, Synthetic Users can maintain context across multiple interactions, enabling deeper and more consistent engagement.

How do you measure Synthetic Organic Parity?

We’re very open about how we measure it and improve on it. The latest scientific research looking at LLMs as Human emulators.