Every day, your brain makes thousands of decisions under uncertainty. Most of the time, you guess right. When you don't, you learn. But when the brain's ability to judge context or assign meaning falters, thoughts and behavior can go astray. In psychiatric disorders ranging from attention-deficit/hyperactivity disorder to schizophrenia, the brain may misjudge how much evidence to gather before acting—or fail to adjust when the rules of the world change based on new information.
"Uncertainty is built into the brain's wiring," says Michael Halassa , a professor of neuroscience at Tufts University School of Medicine . "Picture groups of neurons casting votes—some optimistic, some pessimistic. Your decisions reflect the average." When that balance skews, the brain can misread the world: assigning too much meaning to random events, as in schizophrenia , or becoming stuck in rigid patterns, as in obsessive-compulsive disorder.
Understanding those misfires has long challenged scientists, says Halassa. "The brain speaks the language of single neurons. But fMRI—the tool we use to study brain activity in people—tracks blood flow, not the electrical chatter of individual brain cells."
Bridging that gap means combining insights from single-cell studies in animals, human brain imaging, and behavior. Now, a new kind of computer model—grounded in real biology—lets researchers simulate how brain circuits make decisions and adapt when the rules change.
Called CogLinks, the model builds biological realism into its design, mirroring how real brain cells are connected and coding for how they assign value to often ambiguous and incomplete observations about the external environment. Unlike many artificial intelligence systems that act like "black boxes," CogLinks shows researchers exactly how its virtual neurons link structure to function. As a result, scientists can map how this virtual brain learns from experience and pivots based on new information.
In a study published October 16 in Nature Communications, senior author Halassa and colleagues at Massachusetts Institute of Technology (MIT) used CogLinks to explore how brain circuits coordinate flexible thinking. Like a flight simulator for the brain, CogLinks let the researchers test what happens when key decision-making circuits go off course. When they weakened the virtual connection between two simulated brain regions—the prefrontal cortex and the mediodorsal thalamus—the system defaulted to slower, habit-driven learning. That result suggests this pathway is essential for adaptability.
To see if those predictions held true in people, the team then conducted a companion fMRI study , which was supervised by both Burkhard Pleger from the Ruhr-University Bochum and Halassa. Volunteers played a game in which the rules unexpectedly changed. As expected, the prefrontal cortex handled planning and the deep, central region of the brain known as the striatum guided habits—but the mediodorsal thalamus lit up when players realized the rules had shifted and adjusted their strategy.
The imaging confirmed what the model had forecast: the mediodorsal thalamus acts as a switchboard linking the brain's two main learning systems—flexible and habitual—helping the brain infer when context has changed and switch strategies accordingly.
Halassa hopes the research helps lay the groundwork for a new kind of " algorithmic psychiatry ," in which computer models help reveal how mental illness emerges from changes in brain circuits, identifying biological markers to precisely target treatments.
"One of the big questions in psychiatry is how to connect what we know about genetics to cognitive symptoms," says Mien Brabeeba Wang , the lead author of the CogLinks study, a co-author of the fMRI study, and an MIT doctoral student in Halassa's lab.
"Many schizophrenia-linked mutations affect chemical receptors found throughout the brain," says Wang. "Future uses of CogLinks may help us see how those widespread molecular changes could make it harder for the brain to organize information for flexible thinking."
Research reported in the CogLinks study was supported by the National Institutes of Health's National Institute of Mental Health under grants P50MH132642, R01MH134466, and R01MH120118 and by the National Science Foundation under grants CCR-2139936, CCR-2003830, and CCF-1810758. Bin A. Wang of South China Normal University served as lead author on the fMRI study. The fMRI study was supported by the National Natural Science Foundation of China; Research Center for Brain Cognition and Human Development, Guandong, China; Guangdong Basic and Applied Basic Research Foundation; Deutsche Forschungsgemeinschaft (DFG, German Research Foundation); and the FoRUM grant. Complete information on authors, funders, methodology, limitations, and conflicts of interest is available in the published paper. The content is solely the responsibility of the authors and does not necessarily represent the official views of the funders.
