Research

Experiments, benchmarks, and technical papers from Minds Research Lab on synthetic respondents, persona-conditioned AI, evaluation methods, and the limits of simulated evidence.

2026-11-01
Introducing Minds Mosaic: Pacing the frontier of audience simulation
Minds Mosaic, our frontier model for audience simulation, predicts how a population answers survey questions before the survey runs. How it scores, category by category, against published results.
2026-10-22
SINUS Milieu-Minds in a Reality Test Against a SINUS Survey
200 SINUS Milieu-Minds answered 24 questions on personal concerns. Per milieu, their answer distributions reached 91.7% closeness to a SINUS survey of 2,005 people, against 96.3% between two real survey waves.
2026-09-19
Do Minds rank Super Bowl ads like real people?
100 grounded Minds rated all 54 ads in the USA TODAY Ad Meter 2026. Their ranking closely matched the human panel's (Spearman 0.85), but they rated the ads about 4% of the scale too kindly.
2026-09-05
ANES 2024: Comparing Synthetic and Later Human Answers
An ANES pre/post-election comparison tests 300 matched people on 28 later answers, with separate results for response format and participant grounding.
2026-09-05
CES 2024: Survey Fit Across Multiple Question Forms
A CES comparison evaluates 300 matched respondents on 27 binary, ordinal and multiselect questions, separating aggregate fit from individual prediction.
2026-09-05
GSS 2024: Preserving Uncertainty Improves Survey Fit
A target-separated GSS pilot compares four response conditions across 360 people and 17 questions, separating probability elicitation from profile value.
2026-09-05
Held-Out Interviews: When Participant Memory Helps
Two public interview corpora test whether compact profiles and retrieved earlier evidence improve fidelity to 66 later answers from 22 real participants.
2026-09-05
PISA 2022: Synthetic Survey Fit Across Five Countries
A five-country PISA comparison evaluates 600 students and ten held-out questions, with weighted distribution errors and explicit country and language limits.
2026-09-05
PRISM Alignment: Testing Participant Fit in Held-Out Answers
A 299-person, 869-item comparison measures whether grounded Minds better reflect participants' values and reasons, with a blinded judge and clear task limits.
2026-09-05
When Synthetic Audiences Outperform Generic Foundation Models
A held-out interview comparison tests when participant profiles and memory improve synthetic responses over generic GPT, Claude and Gemini prompts.
2026-08-21
Reference: Minds Research Method Pipelines
Product reference for supported Minds research-method workflows, their inputs, calculation stages, output artifacts, and limits where enabled.
2026-08-14
We Tested Synthetic Audiences Against Reality
What five public survey datasets reveal about synthetic audience accuracy, from a 301-Mind Gen Z test to an international PISA replication, and where grounding helps.
2026-08-14
Can AI Recreate a Real Gen Z Food Survey? Minds Reached 93.99% Approximation
An outcome-blind validation compared 903 responses from 301 persistent Gen Z Minds with published UK Food Standards Agency survey distributions. Aggregate approximation reached 93.99%.
2026-07-26
Audience Grounding Reference: Data Sources
A reference list of public statistical, academic, and industry sources that may support a Minds Audience when relevant, accessible, and permitted.
2026-07-10
Minds PRISM: The AI Engine Behind Synthetic Research
Minds PRISM is the reasoning, inference and source-modeling engine beneath every Mind. See how it grounds AI personas, builds Audiences and where it stops.
2026-07-03
Practical Guide: Before You Trust a Synthetic Audience
A practical review guide for Audience definition, grounding, prompt neutrality, output inspection, reporting, and follow-up evidence.
2026-06-21
Inside Minds: How Synthetic Research Panels Are Built
How Minds creates reusable AI-persona Audiences, collects parallel synthetic responses, and fits into evidence-conscious research workflows.
2025-10-17
Persona Conditioning Reduced AI Output Convergence
In seven creative tasks, persona-conditioned agents produced more diverse outputs than a uniform baseline, evidence relevant to differentiated synthetic research responses.

Introducing Minds Mosaic: Pacing the frontier of audience simulation

SINUS Milieu-Minds in a Reality Test Against a SINUS Survey

Do Minds rank Super Bowl ads like real people?

ANES 2024: Comparing Synthetic and Later Human Answers

CES 2024: Survey Fit Across Multiple Question Forms

GSS 2024: Preserving Uncertainty Improves Survey Fit

Held-Out Interviews: When Participant Memory Helps

PISA 2022: Synthetic Survey Fit Across Five Countries

PRISM Alignment: Testing Participant Fit in Held-Out Answers

When Synthetic Audiences Outperform Generic Foundation Models

Reference: Minds Research Method Pipelines

We Tested Synthetic Audiences Against Reality

Can AI Recreate a Real Gen Z Food Survey? Minds Reached 93.99% Approximation

Audience Grounding Reference: Data Sources

Minds PRISM: The AI Engine Behind Synthetic Research

Practical Guide: Before You Trust a Synthetic Audience

Inside Minds: How Synthetic Research Panels Are Built

Persona Conditioning Reduced AI Output Convergence