Human–AI collaboration · Symbiotic Architecture
The promise was simple: a person plus an AI outperform either one alone. The data said otherwise.
In a broad meta-analysis —106 studies—, human–AI combinations performed, on average, below the better of
their two components working alone.
The problem isn't the AI. It's that combining a person and a machine is not the same as designing how
they work together. And when it isn't designed, one of two things usually happens: either the person
ignores what the AI knows, or surrenders to what the AI says. In both, someone stopped thinking.
This is an essay on how to design collaboration between people and artificial intelligence. The finding from the 106 studies is published evidence; the why —combining without designing—, the three ways of working with AI and the five principles are the author's proposal —the subject of the book Arquitectura Simbiótica—, not a result of the study.
Study: Vaccaro, Almaatouq & Malone · Nature Human Behaviour · 2024. Framework: Arquitectura Simbiótica (2nd ed., 2026).
For years it was assumed that adding AI to a task improves it. Vaccaro and colleagues' review —106 experimental studies— found an uncomfortable nuance (below). The cause wasn't that the AI was bad —often it was more accurate—, but that the way they were combined wasted what each did best.
The study measured combinations, not designs. This essay's wager —a hypothesis, not a result of the meta-analysis— is that the difference lies in which of the three ways you use.
They look alike from outside and give opposite results. Pick one and watch what happens to your judgment in each case.
That the result is the good one —the one that expands you instead of dimming you— isn't luck. It's design, and it rests on five principles. The last governs the other four.
The full method
This is the idea. The blueprint is in the book.
The three forms, the five principles, the models and a worked case from start to finish —how to design symbiosis instead of leaving it to chance— are the subject of Arquitectura Simbiótica, second edition (written in Spanish). Written from more than two decades designing mission-critical systems for government, for anyone who works with AI, decides about it, or simply wants to understand what changes when AI stops being a tool.
Explore the book · arquitecturasimbiotica.co →What it is. An essay on designing collaboration between people and artificial intelligence. It starts from an empirical finding —Vaccaro et al.'s (2024) meta-analysis— and offers a framework to understand it and act on it.
What is data and what is proposal. The figure of 106 studies and its result are published evidence, cited with its source. The three ways of working with AI and the five principles are a proposed framework —that of the book Arquitectura Simbiótica—, not a measurement or an established consensus.
Central study: Vaccaro, M., Almaatouq, A. & Malone, T. (2024). «When combinations of humans and AI are useful: A systematic review and meta-analysis». Nature Human Behaviour. Conceptual framework: Navarrete Ruiz, C. A. (2026). Arquitectura Simbiótica (2nd ed.). ISBN 978-628-02-4845-5. Own work.