The four research methods, on a population that persists.
Survey, interview, intervention and experiment on the same personas, with attitudes and memory you can inspect and a methodology page that is upfront about the limits.
What gets in the way.
One-off tools
Most AI research tools discard the population after each question.
Opaque methods
It is hard to defend output when you cannot say how it was produced.
No clean controls
Comparing treatments requires equal groups.
How researchers use Artixol.
Define the sample
Brief, culture, optional places and relationships.
Run the instruments
Survey, Ask, Shock, Message, Rumor.
Inspect the people
Person cards show attitudes and memories.
Export and cite
CSV and JSON for analysis, with limits stated.
Use cases for researchers.
User research surveys
Ask everyone the same questions, then read the table and the splits by role.
Learn more → LabA/B testing on the same people
Run two treatments on clones of the same named people so the comparison is clean.
Learn more → ShockPolicy and shock testing
Inject a price hike, an outage or a new rule and see who is hit and who is unmoved.
Learn more →Questions from researchers.
Is Artixol a substitute for a representative sample?
No. See our methodology page. Outputs are experimental model output for exploration, not a census.
Can I get the raw data?
Yes. Studies export to CSV and JSON.
Bring a question. We will staff the audience.
Tell us who you need to understand. We will build the audience and run a study with you.