Centaur, the AI that understands the human mind: a new era for psychology, medicine and digital ethics

A new artificial intelligence model, called Centaur, could revolutionize the way we understand human behavior. Created by an international team of researchers and recently presented in the prestigious journal Nature, Centaur is the first AI system to accurately predict people's decisions in almost any psychological experiment - surpassing traditional cognitive models and offering new insights into how the human brain works.

• What is Centaur?

Centaur is an AI model trained on data from over 60,000 people and 10 million decisions. Instead of specializing in a single type of task, it learns broadly - from diverse experiments, from memory tests to moral dilemmas - and manages to accurately anticipate human reactions in completely new, previously unseen contexts. The model was built on Llama 3.1, a language model developed by Meta, and was trained in record time: just five days, with minimal modification of the initial parameters. The key to success was the Psych-101 database, a set of 160 psychological experiments described textually to be intelligible to an AI.

• A mirror of human cognition

The results were spectacular: Centaur not only beat classical models, but also demonstrated the ability to simulate realistic human behavior, even in abstract tasks or in completely changed situations. More amazingly, the researchers found that the internal structure of the model begins to resemble human brain activity. Comparing the internal states of the AI with brain scans of human subjects, stronger correlations were observed than in the case of any previous model. In other words, Centaur does not just predict what we do, it seems to "think” in a way that is close to how we think.

• Implications for medicine, psychology and science

This achievement has the potential to profoundly change several fields: Psychology and mental health: it could become a revolutionary tool in diagnosing and personalizing therapies; Medicine: it offers a window into understanding cognitive processes, useful in neurology and cognitive recovery; Education: it models learning methods adapted to the cognitive style of each student; Basic science: Centaur has even identified a new decision-making strategy, unknown until now, which turns it into an active ally of scientific discovery.

• A fascinating technology

The ability to anticipate people's thinking and decisions raises major ethical questions. What happens if such models are used for commercial, political or military purposes? How do we protect privacy, given that AI can extrapolate behaviors from seemingly trivial choices? Researchers warn that this technology must be strictly regulated. Although Centaur is not built for manipulation, its ability to understand human thought can easily be hijacked if there is no solid ethical framework.

• Next Steps

The model is still in its early stages: it mainly covers learning and decision-making, but not essential areas such as social psychology or cultural differences. Also, most of the data comes from Western backgrounds, a common imbalance in psychological research. The team aims to create a unified cognitive model, including diverse populations and cultural contexts. Centaur and its dataset are already public, so that other researchers can develop them further. Centaur marks the beginning of a new era: one in which we can not only observe human thought, but also simulate, predict, and - perhaps - understand it better than ever before. It remains to be seen whether it will help us know ourselves more deeply or whether it will open a Pandora's box of cognitive manipulation. One thing is certain: the future of psychology and AI is being written, right now, in common language.