Mark Twain said that humans are the only animal that blushes. Or what you need to do it. But this is not what defines us as a species. Is intelligence? Culture? For some scientists, the answer is in our decisions, the choices we take before each bifurcation of life.
With this in mind, a team of scientists from the Helmholtz Munich Institute has developed A new language model that simulates human behavior with amazing precision.
Called Centauro, in honor of the mythological creature, the AI model It has trained with more than ten million individual decisions made by more than 60,000 participants in 160 psychological experiments.
Thanks to this it reflects How people think and make decisions both in family and unknown situationsopening new ways to understand human cognition and improving psychological theories.
For decades, cognitive science has worked for Build models that can explain and predict human behavior. Traditional approaches have generally achieved one or the other, but not both. Centaur changes that.
The team, led by Marcel Binz and Eric Schulz, has resorted to a carefully selected data set called Psych-101, which captures a wide range of behavior patternsincluding learning based on risks and rewards, and moral reasoning, and structure in an understandable format for a language model.
Unlike the previous models that are based on predefined rules or narrow parameters, Centauro learns common decision -making strategies and generalizes them to new contexts. It even predicts reaction times, offering a deeper vision of the dynamics of the choice.
“We have created a tool that It allows us to predict human behavior in any situation described in natural language, As a virtual laboratory, ”explains Binz in A statement.
According to the results, published in Nature, these capacities could support clinical research by helping Simulate the decision -making processes of people with mental health problems, such as anxiety and depression. Beyond psychology, the model could improve the modeling of decision -making in medical care, policy design and social sciences.
The next phase of the investigation will focus on two sections. The first will be to examine the internal functioning of Centaur. The team seeks to identify which computational patterns correspond to specific decision -making processes and if these can provide information on how people process information.
And the second objective is Expand the samples of the answers to include other demographic and psychological characteristics. A key question is whether the model can help distinguish between cognitive strategies used by healthy people and those with mental health problems.
Binz’s team believes that These models have the potential to significantly improve our understanding of human cognitionprovided they are applied in a responsible manner.
“We combine research in AI with psychological theory, with a clear ethical commitment -concludes Binz -. In a public research environment, we have the freedom to address fundamental cognitive issues that are often not the focus of industry care ”.