Understand people before you speak to them.
Impulse Labs is a Mexico City lab that brings together artificial intelligence, audience analysis and psychology. We use them to build tools for communication teams, books and explainer content, and self-knowledge products.
Hybrid Intelligence: one hundred simulated people read a sample message
Fictional example: a city announces roadwork on a main avenue.
“The avenue will close for three months for construction.”“For three months there will be detours; once the work is done, your commute will be shorter.”
- Convinced
- No clear position
- Has an objection
Each square is a simulated person; the color shows their likely reaction. An illustration of the kind of reading, not a measured result.
Lines of business
-
Hybrid Intelligence
Simulated people, anchored in Mexico’s 2020 Census and in field data from INRA, read a message before it goes out. For government, brands, agencies and media.
Explore the method -
Digital publishing
Books, guides and explainer videos on psychology, philosophy, faith, business and technology, in Spanish, English and Portuguese.
See our publishing line -
Ethea
Self-knowledge and well-being through science-based tests and training, in four languages.
Go to ethea.me -
Open Lab
The lab’s technical foundations, open source and built on public INEGI data.
Explore the Open Lab
Hybrid Intelligence
Test your message with a simulated audience before you publish it. Each simulated person reads the message in the context of the day and tells you what convinces them, what they don’t understand and what bothers them.
How a reading comes together
A sample anchored in the Census
Municipalities are drawn in proportion to their adult population, and each profile takes its traits from the 2020 Census. Design weights and effective sample size are reported.
A real person behind every profile
Each profile’s daily life and media habits come from INRA field interviews, protected with k-anonymity. Then each person reads your message alongside the day’s news.
A reading that states its limits
Paired A/B comparison with confidence intervals, an A/A control and a fixed seed. Every run comes with a method sheet and a list of what the simulation can’t tell you.
Who we work with
One Hybrid Intelligence core, with tools built for each kind of team.
KAPITALRigor by design
- A/B comparison paired by person, with a 95% confidence interval and a minimum detectable effect.
- An A/A control to measure the noise in each run.
- Answer order reversed in half the batches, to counter the models’ position bias.
- Fixed seeds and a method sheet for every run: the same question gives the same sample.
We have not yet measured how faithfully the method reflects Mexico’s population. We are preparing an open study that compares it with official INEGI surveys, with forecasts registered before each release. See the validation harness in the Open Lab
What it is not
- It does not measure public opinion.
- It does not estimate voting intention.
- It does not forecast sales or ratings.
Use it to anticipate objections, compare wording and form hypotheses before you invest in a message.
Want to see a reading of your own message?
We’ll walk you through the tool with a case from your sector in a 30-minute call.
Request a demo