Search papers, labs, and topics across Lattice.
Anamnesis is an open-source platform that leverages large language models to simulate surveys with controlled demographics, allowing researchers to prototype and stress-test survey instruments without involving real human subjects. By integrating the Anthology and Alterity frameworks, it conditions model responses using structured narrative backstories, facilitating open-ended generation and multimodal surveys. Evaluations show that Anamnesis yields opinion distributions that align more closely with actual survey data compared to traditional persona-prompting methods, highlighting its potential as a robust alternative to proprietary tools.
Anamnesis produces survey results that mirror real-world data more accurately than conventional methods, revolutionizing how we simulate public opinion.
We present Anamnesis, an interactive system for demographically controllable survey simulation using large language models. Open-source, and designed for non-technical users/researchers, Anamnesis enables the prototyping and stress-testing of survey instruments on virtual populations rather than real human subjects. The platform operationalizes the recently introduced Anthology and Alterity frameworks, which use structured narrative backstories to condition model responses, within a unified web interface. It supports open-ended generation, probabilistic demographic resampling, and multimodal (image and audio) surveys. We evaluate the system through two case studies: (1) replicating segments of Pew Research Center's American Trends Panel (ATP) on political typology and biomedical issues and (2) emulating human preference in the New Yorker Caption Contest. In both cases, Anamnesis produces opinion distributions that more closely match real-world survey data than standard persona-prompting baselines, offering a transparent, reproducible, and open-source alternative to proprietary simulation services.
