A traditional poll or focus group has a feedback loop of weeks and months, and cost tens of thousands of dollars. What if you could get similar results in a matter of minutes at a fraction of the cost?
The objective isn't to perfectly replicate human behavior. Instead, it's about aligning your synthetic audience with real-world patterns to create a powerful feedback loop.
What sets our approach apart is how we spin up individual instances of each model to role-play as if it were a real person.
We're actively integrating with our marketing platform, which connects to third-party data sources like Meta Ads. This integration enables us to create more sophisticated focus groups and run surveys that closely mirror your real-world target audience.
These are the features we’re gearing up to release soon.
2024
Imagine LLMs stepping into human shoes. How do they fare in classic behavioral economics experiments? This groundbreaking MIT & Harvard study reveals that LLMs mirror human behavior in auctions, job interviews, and negotiations. The findings suggest a promising future for using LLMs in polling and surveys, potentially revolutionizing how we gather and analyze data on human behavior.
2023
What are some of the practical ways you might apply LLMs to market research? This paper explores that by focusing on consumer behavior and buying habits. You'll see how LLM personas respond to consumer behavior and behavorial economics experiments in the same way as humans do. These findings strongly support the potential of synthetic focus groups and polling in market research or understanding voter preferences.
2024
Listen to Anthropic CEO and Co-founder Dario Amodei give the best explanation of how synthetic data works that we've heard. He clearly and simply explains the process that Google's Deep Mind used to train an AI to learn the game of Go, and how that is an example of training a model on synthetic data.
2023
This is a comprehensive report summarizing how LLMs and agentic workflows are reshaping various aspects of work. It provides valuable insights into the future of research and an AI-driven workplace.
2024
This is a really interesting experiment conducted by lawyer and legal blogger Adam Unikosky where he shows that Anthropic's Claude 3.5 is capable of adjuducating Supreme Court cases in a manner comparable to the Justices themselves.
Adam Unikosky's Legal Newsletter
2023
There is latent knowledge about human behavior latent in LLMs waiting to be unlocked. This study by MIT's John Horton is one of the first to explore this idea. By comparing LLM responses to experiments with human subjects, it reveals the understanding of human behavior embedded in these models, and lays the foundation for synthetic behavioral research.
2022
This is one of the first studies of its kind that shows that LLMs, when role-playing as humans, respond in the same way as humans might by correlating the results of LLM reponses with famous studies in the social sciences. It suggests that we can run experiments about human behavior "in silica." It points the way to the synthetic focus group technique we can now deploy today.
