About Expected Parrot
Expected Parrot enables researchers and teams to study human responses using AI simulations and survey research. The goal is to produce results that are structured, reproducible, and usable as part of real research workflows.
Why Expected Parrot?
User research and social science experiments are expensive, slow, and difficult to iterate on. Large language models make it possible to simulate how people might respond in surveys or interviews. But most AI tools produce unstructured answers that are difficult to compare, analyze, or validate.
Expected Parrot was built to make AI-driven research reproducible and testable. Simulations generate datasets that can be analyzed, compared, and reused as part of a study.
Why open source?
Work that claims to represent human behavior must be transparent in order to be defensible.
When AI systems generate responses meant to stand in for people, the methods behind those results matter. Researchers and teams need to see how agents are defined, how surveys are constructed, how models are used, and how results are produced. Those methods must be visible and reproducible.
Foundational research
Expected Parrot builds on early research into language models as simulated social agents led by co-founder John Horton, a professor at MIT Sloan. The company was founded by John Horton and Robin Horton.
Large Language Models as Simulated Economic Agents: What Can We Learn from Homo Silicus?
John J. Horton - January 2023
Introduces the idea of treating language models as simulated economic agents that can be placed into experimental scenarios to study behavior.
Automated Social Science: Language Models as Scientist and Subjects
Benjamin S. Manning, Kehang Zhu, John J. Horton - April 2024
Explores how language models can both generate social-science hypotheses and act as experimental subjects in computational studies.
Benjamin S. Manning, John J. Horton - September 2025
Develops methods for constructing agents that can generalize across scenarios and simulate behavior in new experimental settings.
Research using Expected Parrot
A growing number of papers and projects use Expected Parrot for simulations, experiments, and survey-based research.
Who We Serve
From universities and research labs to startups and global enterprises, Expected Parrot is built for teams that care about evidence—supporting testing, learning, and iteration with confidence.
Supporters
Expected Parrot is part of Y Combinator (Fall 2025) and backed by Bloomberg Beta and other investors, alongside a growing global community of researchers and builders.