Our research methodology

Six steps from raw census data to structured, actionable research.

Data Foundation
01
Real-world data grounding
Census indicatorsPopulation pyramidsWeb search
02
Demographic stratification
Age · Gender · IncomeUrbanization · EducationPer-stratum quota planning

Every panel starts with real-world census indicators, population pyramids, and live web searches. The raw data is split into strata across age, gender, income, urbanization, and education, each with a target quota that mirrors real-world distributions.

Strata & target distributions
AI Persona Construction
03
Largest-remainder allocation
Integer optimizationAggregate reconciliation
diversity constraints
04
Verbalized sampling
OCEAN personality candidatesTypicality scoringDiversity-aware selection

A largest-remainder algorithm allocates integer headcounts so every quota is honoured exactly. For each slot, 3–5 personality candidates are scored on the OCEAN model, filtered for typicality, and selected for diversity—producing a broad, realistic spectrum of AI-simulated attitudes.

AI-simulated personas
Research Execution
05
Parallel independent interviews
Isolated contextNative-language dialogueMulti-granularity categorization
06
Real-time fact verification
Source-grounded checkingVerified / flagged claims

Each AI persona is interviewed in its own isolated context, in its native language, by a dedicated AI interviewer. Major factual claims are checked in real time via automated web search and labeled as verified or flagged with source attribution.

See it in action

Run your first research session in under a minute.

Get started