What is is an independent research lab exploring new mediums of thought and expanding the imaginative powers of the human species. We have created a tool that allows people to conduct causal experiments, at scale. Causal experiments are frequently used in product design, transportation, political science, public health, and economics when decision-makers want to understand how a population will respond to different options. For instance, researchers around the world performed experiments regarding attitudes toward different aspects of COVID-19 vaccines to understand what could be done to encourage vaccine acceptance before vaccines actually became available. is able to dramatically reduce the cost (10,000x less costly and 1,000x less time) needed to conduct such surveys. Furthermore, our techniques eliminate concerns about self-selection bias and oversampling. If you’d like to learn more, please reach out to our commercial team.

How do people use

As of February 2023, is a brand-new technology. Our users are primarily researchers (students and researchers in academia, at independent organizations, or operating independently) trying to understand how populations make decisions. makes professional-level research available for everyone from middle school students to policymakers to answer questions about why people make the decisions they do. With the workflow, you can:

  • Ask any question about human behavior.
  • Focus on an outcome (or dependent variable) that the behavior affects, such as increasing value or reducing risk.
  • Design your experiment with the guidance of AI to select the attributes to study and levels that will vary in your survey.
  • Survey a population using your experiment and get analyzed results to understand the causal effects of decisions.
  • Take a deeper look to see how population characteristics affect the decisions made. is an early-stage product with frequent updates and improvements.

What are the limitations of

To help you calibrate how much you can rely on, we’ll share some of the limitations you should be aware of as you use our technology:

Limitations specific to

Currently, implements limits to the number of attributes and levels that can be examined in a single experiment to maintain speed and computational efficiency. While there is no theoretical limit to the number of attributes and levels that can be tested, these must be limited for cost and human attention span. We implement the standard analysis methodologies at this time, specifically OLS and CLM regressions. Researchers wishing to try other analyses can download any survey data directly.

Limitations that apply to all tools using large language models is only as good as the data upon which it was trained. If the respondents are biased, that may show up in survey responses. We have compiled results from a number of published studies using similar methodologies and compared them to our results. These comparisons have been favorable, showing no more difference between the AI-generated results and the results of a published study than are seen when data from a published study was split into three parts and compared against itself. However, if the question asked of is not related to the types of published studies we were able to evaluate, we cannot know for sure the results would be similar. In short, there is no guarantee of our results!

Other thoughts on limitations

This section is still too short. We tried to share enough to make you not over-rely on, but this is not a comprehensive list of possible limitations.

Who is building

We are a small self-funded team focused on design, human infrastructure, and AI. We have several full-time staff and an incredible set of advisors.

What makes more ethical than standard research experiments?
  • Consent and privacy: LLMs and Synthetic Respondents do not have personal feelings, emotions, or privacy concerns. Therefore, researchers do not need to worry about obtaining informed consent or protecting the privacy of their participants. In contrast, involving human subjects requires researchers to address these ethical considerations, which can sometimes be challenging and time-consuming.
  • Minimizing harm: Experimenting with LLMs and Synthetic Respondents eliminates the risk of causing psychological or emotional distress to human subjects. Some experiments may involve sensitive topics or require participants to experience discomfort, which raises ethical concerns. By using LLMs and Synthetic Respondents, researchers can ensure that no harm is done to any sentient beings.
  • Resource efficiency: Conducting experiments with LLMs and Synthetic Respondents can be more cost-effective and less time-consuming than recruiting, training, and compensating human participants. This increased efficiency allows researchers to allocate resources to other aspects of their work, ultimately advancing knowledge more quickly.
  • Flexibility: LLMs and Synthetic Respondents can participate in a wide range of experiments, including those that may be logistically difficult or impossible to conduct with human subjects. This flexibility allows researchers to explore new and innovative research questions.

However, it is important to recognize that the use of LLMs and Synthetic Respondents has its limitations and may not be appropriate for all types of experiments. For instance, when studying complex human emotions, social interactions, or decision-making processes, real human subjects may still be necessary to gain accurate insights. Additionally, it is crucial to ensure that the LLMs and Synthetic Respondents used in research are designed and trained ethically and do not perpetuate biases or misinformation.

What do I do if I have a problem?

You can post a message in the #conjoint-help channel on the Discord server. Please give as detailed a description as possible of any error, including the error message you encountered. Screenshots and screen recordings of the problem also help.

How do I submit a feature request?

Post a message in the #feature-requests channel in the Discord server.

How can I share feedback?

Post a message in the #general channel in the Discord server. We find your feedback valuable. Please share.

How can I learn more?

Join the Discord server!

How can I help?

Thank you for your willingness to help! Here are some ways to contribute: You can join our team, or refer someone who ends up joining our team. If you’d like to work part-time as a research assistant, please send an email to If you are a researcher you can evaluate our workflow. Post a message on the Discord server to learn more.