An international research team has presented “Centaur”, an artificial intelligence model that can predict human behavior in a wide range of cognitive tasks with incredible accuracy.
This is the result of the study “A foundation model to predict and capture human cognition”, which published in Nature.
Centaur uses Meta AI’s Llama 3.1 (70B) language model and was optimized with a specially curated dataset called “Psych-101.” This includes over 10 million decisions from more than 60.000 participants in 160 psychological experiments in areas such as object classification and more.
The researchers point out that, although the dataset is very large, it is not perfect: for example, information on the age or socioeconomic status of the participants is missing, and the test participants come mainly from Western countries.
The team trained Centaur with 90% of the data and then tested how well the model could predict decisions made by people who were not included in the training dataset.
Even with slightly different testing settings, the predictive power of the model remained high. In both cases, Centaur performed better than the original Llama model or other models.
How Centaur and Psych-101 work
Psych-101 includes data from 160 psychological experiments with 60.092 participants who made a total of 10.681.650 decisions and generated 253.597.411 text tokens. The dataset includes areas such as decision-making, memory, supervised learning, and others.
Centaur is a basic model of human cognition that has been optimized in Psych-101, among other things.
Information processing mechanisms in humans?
The researchers also show that the model's internal workings correlate closely with measured brain activity (fMRI data) – even though it was not trained to do so. This suggests that Centaur uses similar information-processing mechanisms to the human brain.
In the future, Centaur could help develop psychological theories, simulate experiments, or better understand individual decision-making strategies.
According to the researchers, both the Psych-101 dataset and the Centaur model could be widely used in behavioral research in the future, for example to prepare new experiments or to better understand cognitive processes. The model also offers new possibilities for simulating the behavior of research participants.
The authors speak of a “tipping point for cognitive science.” The Llama-3.1-Centaur-70B model and the “Psych-101” dataset are available at Hugging Face.
The team plans to further expand the dataset – to include psycholinguistic, social, and cross-cultural aspects of human behavior.
“Not yet ready for complex everyday situations”
Experts such as Professor Andreas Glöckner from the University of Cologne, Professor Markus Langer from the University of Freiburg and Professor Clemens Stachl from the University of St. Gallen see Centaur as an important step forward.
However, they also emphasize that the predictive power has so far been limited to laboratory and decision-making experiments and cannot yet be transferred to complex everyday situations.
They also warn against irresponsible interpretation of the results and urge caution regarding ethical issues and potential misuse. Langer questions whether the “accuracy in predicting human behavior – which in some cases is 64% – can really be rated as ‘good’.”
DOI: https://doi.org/10.1038/s41586-025-09215-4
Although the press releases will range from very select to rare, I said I'd pass...because sometimes the editors hide.
