how network structure shapes decision-making

Network
how network structure shapes decision-making

Do intelligent people think faster than others when solving problems? The results of a new study by Human Brain Project researchers at Charité University Berlin together with their collaborators at University Pompeu Fabra in Barcelona, ​​published in Nature Communications, are challenging this long-held assumption in intelligence research. Taking a biologically inspired approach, they built 650 personalized brain network models (BNMs) based on data from the Human Connectome Project and simulated the brain dynamics involved in problem solving.

how network structure shapes decision-making

Observations from the brain simulations were compared to empirical data of the 650 participants taking the so-called Penn Matrix Reasoning Test (PMAT), consisting of a series of increasingly difficult pattern matching tasks. The results of these were quantified into participants’ fluid intelligence (FI), which could roughly be described as the ability to take difficult decisions in new situations.

“We found that people scoring higher on fluid intelligence (FI) took more time to solve the more difficult tasks compared to people with lower FI. They were only quicker when responding to simple questions,” explained Petra Ritter of Charité University, senior author of the study. “We first observed this in our simulations, and then only afterwards did we see that the empirical data of participants taking the intelligence tests corresponded to this trend.” Ritter’s lab and many other research groups at HBP use brain simulations to complement observational data, in order to develop a theoretical framework of how the brain works.

In this case, brain simulation has been employed to determine the link between functional and structural connectivity in the brain and cognitive performance. A more synchronized brain is better at solving problems, but not necessarily faster. “As synchronization is reduced, decision-making circuits in the brain jump faster to conclusions, while higher synchronization between brain regions allows for better integration of evidence and more robust working memory,” says Ritter. “Intuitively this is not so surprising: if you have more time and consider more evidence, you invest more in problem solving and come up with better solutions. Here we do not only show this empirically, but we demonstrate how the observed performance differences are a consequence of the dynamic principles in personalized brain network models. We thus present new evidence that challenges a common notion about human intelligence.”

Previously established local circuit models of working memory (WM) and decision-making (DM), both important for intelligence, were plugged into The Virtual Brain (TVB), of which the latter provided a simulation at the whole-brain level.

The simulations were run using a multi-scale brain modeling approach; brain imaging data were processed with automated containerized pipelines. The processing of the highly sensitive brain data took place within a secure Virtual Research Environment of EBRAINS Health Data Cloud. These technologies are accessible though EBRAINS to the global research community.

The ultimate goal of the study is not to find out how fast you should think, but rather to understand how biological networks determine decision-making for the development of bio-inspired tools and robotic applications. Modeling brain dynamics of intelligent decision-making is therefore a promising approach to building smart applications. “We think that biologically more realistic models may outperform classical AI in the future,” says Ritter.

Text: Matthijs de Boer, Roberto Inchingolo

References: Learning how network structure shapes decision-making for bio-inspired computing, Michael Schirner, Gustavo Deco & Petra Ritter Nature Communications volume 14, Article number: 2963 (2023)

Contact media

Peter Zekert
[email protected]

Related Posts