Your Emotions Are Just Data, and They Have You Perfectly Mapped

View on LinkedIn

Imagine someone could predict with mathematical precision what decision you will make before you know it yourself. We are not talking about magic or fortune-telling, but about artificial intelligence trained on millions of real human decisions.

That technology already exists. It is called Centaur, and it represents a quantum leap in machines' capacity to understand — and potentially manipulate — human behavior.

The Construction of Humanity's Digital Mirror

Centaur is no ordinary AI. Researchers at the Helmholtz Institute in Munich took LLaMA, Meta's language model, and subjected it to an unprecedented fine-tuning process. Over five intensive days, they fed the model more than 10 million real human decisions drawn from 160 psychological experiments involving 60,000 participants.

The process was meticulous: they presented the AI with the same scenarios that humans faced in the experiments — moral dilemmas, memory games, economic decisions — and trained it to predict what each person would choose. When it was wrong, they adjusted the model. Iteration after iteration, Centaur learned not only our logic, but also our contradictions, biases, and irrationalities.

The result is chilling in its precision: Centaur can correctly predict 64% of human decisions and outperformed 14 traditional cognitive models on 31 of 32 tasks. It does not only predict what we will decide, but also how long it will take us to decide.

A Potential Laboratory for Manipulation

What Centaur does goes beyond academic prediction. This AI can simulate how you would react to any situation described in natural language. What price would make you buy that product? What political message would persuade you? What advertisement would make you click?

The "beneficial" applications are obvious: accelerated psychological research, more intuitive product design, consumer behavior prediction. But the line between understanding and manipulation is dangerously thin.

If a company can simulate thousands of versions of yourself to find exactly what combination of words, colors, and emotions will make you buy its product, are you still making a free decision? If a politician can test on Centaur which speech will activate your deepest fears or most intimate hopes, is that democracy or social engineering?

The Elephant in the Room: Who Else Has This Technology?

Here is the most unsettling part: Centaur is the first model of its kind to be made public, but do we really believe it is the first one to exist?

Amazon tracks every click, every pause, every product you look at but do not buy. TikTok analyzes how many milliseconds you stop on each video, what makes you swipe, and what keeps you hooked. Meta knows your reactions, your likes, your messages. Google knows what you search for at 3 AM when you cannot sleep.

Each of these companies has orders of magnitude more data on human behavior than what was used to train Centaur. If academic researchers with limited resources were able to create such a powerful model, what predictive and manipulative capabilities must these corporations have developed in their private laboratories?

The difference is that Centaur is transparent about its capabilities. Corporate AIs operate in the shadows, silently optimizing every pixel, every word, every notification to hack your brain and maximize their objective: your attention, your money, your vote.

The Imminent Danger We Face

We are at an inflection point. The technology to predict and potentially control human behavior at massive scale already exists. The question is not whether it will be used to manipulate us — it already is — but how deep that manipulation will go and whether we will have any defense against it.

Every time you interact with a digital platform, you are feeding models that learn to predict you better. Your emotions become data points. Your "irrational" decisions reveal exploitable patterns. Your humanity is reduced to parameters in an optimization function.

Free will — that comforting illusion that our decisions are truly our own — wavers in the face of evidence that we are surprisingly predictable. And if we are predictable, we are manipulable.

Sources consulted

Academic sources:

Nature: "A foundation model to predict and capture human cognition" Helmholtz Munich: "AI That Thinks Like Us – and Could Help Explain How We Think"

• Nature: "A foundation model to predict and capture human cognition"
• Helmholtz Munich: "AI That Thinks Like Us – and Could Help Explain How We Think"

Science journalism:

Live Science: "New 'Centaur' AI model can predict how we behave with unprecedented accuracy" SciTechDaily: "AI That Thinks Like Us: New Model Predicts Human Decisions With Startling Accuracy"

• Live Science: "New 'Centaur' AI model can predict how we behave with unprecedented accuracy"
• SciTechDaily: "AI That Thinks Like Us: New Model Predicts Human Decisions With Startling Accuracy"

Analysis and opinion:

El País: "Un centauro sueña con pensar" Medium (Robert Encarnacao): "Human Behavior Isn't Random. This AI Proves It — and Predicts Yours!" Ámbito: "Aseguran que una nueva IA puede predecir decisiones humanas" Science (cited in Ámbito): Criticism by Jeffrey Bowers

• El País: "Un centauro sueña con pensar"
• Medium (Robert Encarnacao): "Human Behavior Isn't Random. This AI Proves It — and Predicts Yours!"
• Ámbito: "Aseguran que una nueva IA puede predecir decisiones humanas"
• Science (cited in Ámbito): Criticism by Jeffrey Bowers