Every day, the human brain engages in a complex dance of decision-making, often under conditions of uncertainty. This intricate process involves evaluating countless pieces of information, weighing potential outcomes, and predicting future events. While most decisions made by the brain occur seamlessly and often successfully, there are times when it misfires. When this happens, it can manifest as an inability to accurately judge information, resulting in errant thoughts and behaviors. This phenomenon is particularly evident in a range of psychiatric disorders, such as attention-deficit/hyperactivity disorder (ADHD) and schizophrenia. When the brain fails to gather the appropriate amount of evidence before acting or struggles to adjust its responses to new and changing information, the consequences can be profound.
Michael Halassa, a prominent neuroscientist at Tufts University School of Medicine, sheds light on this complexity. He describes the brain’s inherent uncertainty, likening its neural networks to groups of neurons casting votes—some optimistic and others pessimistic. This interplay of signals forms the basis for our decision-making processes. When this balance tips, the brain can misinterpret reality. For instance, individuals with schizophrenia may assign excessive significance to random events, while those with obsessive-compulsive disorder may become trapped in rigid thought patterns.
Studying these neural misfires has been a daunting challenge for scientists. Halassa emphasizes that the brain communicates through single neurons; however, the primary tool used to study brain activity in humans—functional magnetic resonance imaging (fMRI)—tracks blood flow rather than the electrical signals emitted by individual brain cells. This inherent limitation complicates our understanding of how various circuits influence thought and behavior.
To bridge this gap and develop a more nuanced understanding of brain function, researchers have combined insights from single-cell studies conducted on animals with human imaging data and behavioral analyses. A breakthrough in this research trajectory is a new computational model known as CogLinks. This innovative model is grounded in biological realism and provides an unprecedented look into how brain circuits formulate decisions and adapt when faced with changing rules in their environment.
CogLinks sets itself apart by integrating biological plausibility in its architecture, mirroring the connectivity of biological neurons. Moreover, it encodes how these neurons assess value amid often vague and incomplete external stimuli. Unlike many traditional AI systems, which often operate as “black boxes” obscuring their inner workings, CogLinks offers transparency. Researchers can discern how the digital neurons within the model create a relationship between structural properties and functional outcomes. This visibility allows them to trace how this virtual brain acquires knowledge through experiences and adjusts its responses based on new data.
In a recent study published in the esteemed journal Nature Communications, Halassa and his collaborative team from the Massachusetts Institute of Technology (MIT) harnessed the power of CogLinks to investigate how distinct brain circuits contribute to flexible thinking. The model rouses comparisons to a flight simulator tailored for cognitive processes, allowing researchers to experiment with how critical brain pathways go awry under different scenarios. By deliberately weakening the connection between two simulated brain regions—the prefrontal cortex and the mediodorsal thalamus—the team observed a detrimental shift towards slower and more habitual learning. This result underscores the importance of this neural pathway in facilitating adaptability in thought and behavior.
To validate the predictions established by the CogLinks model in human subjects, the research team conducted a complementary fMRI study, jointly overseen by Burkhard Pleger from Ruhr-University Bochum and Halassa himself. During this experimental phase, participants engaged in a game where the rules occasionally changed without warning. The results aligned perfectly with the model’s projections: the prefrontal cortex was responsible for planning and executing decisions based on established rules, while the striatum, a deep-seated region of the brain, governed habitual responses. Notably, the mediodorsal thalamus became particularly active when players recognized the alteration in rules and adjusted their strategies accordingly.
These imaging results confirmed the role of the mediodorsal thalamus as a pivotal switchboard linking two primary learning systems in the brain—flexible and habitual. This neural coordination is crucial for enabling the brain to recognize when contextual information shifts and to switch strategies to accommodate new circumstances. The implications of this research are far-reaching, providing insight into the fundamental mechanisms underlying decision-making and behavioral adaptation.
Halassa envisions that this groundbreaking research lays the foundation for a next-generation approach to psychiatric treatment, which he conceptualizes as “algorithmic psychiatry.” This framework would utilize advanced computer models to unravel how changes in brain circuits contribute to mental illness, paving the way for the identification of biological markers that can be targeted for more precise treatments.
Mien Brabeeba Wang, the lead author of the CogLinks study and a doctoral student at MIT in Halassa’s lab, underscores the relevance of this research in connecting genetic knowledge to cognitive symptoms associated with psychiatric conditions. Wang elaborates on the significance of this study’s findings, noting that many genetic mutations linked to schizophrenia influence chemical receptors distributed throughout the brain. Future implementations of CogLinks could illuminate how such widespread molecular alterations hinder the brain’s capacity to organize information, ultimately affecting its proficiency in flexible thinking.
The research findings detailed in the CogLinks study received support from several significant grants awarded by the National Institutes of Health’s National Institute of Mental Health, highlighting the collaborative effort invested into elucidating these complex cognitive processes. Additional funding came from the National Science Foundation, enabling the researchers to explore this critical intersection of neuroscience and artificial intelligence.
With groundbreaking advancements in comprehension of brain function being a continual pursuit, emerging technologies like CogLinks stand at the forefront of neuroscience. As researchers strive to uncover the unseen patterns that guide decision-making and behavior, the potential applications of this knowledge will undoubtedly expand, impacting our understanding of mental health and the treatment of psychiatric disorders in meaningful ways.
As research in this field progresses, questions remain about how we can harness the insights derived from advanced computational models to propel innovations in psychiatric treatments and interventions. As the intersection of neuroscience and artificial intelligence continues to evolve, patient outcomes could significantly improve, shifting the paradigm of how mental health challenges are approached and treated.
The CogLinks model represents a remarkable convergence of biological realism and computational prowess, opening new avenues for exploration in understanding both the mechanics of the brain and the complexities of human behavior. As this technology advances, it promises to uncover the intricate details that underpin our thoughts, decisions, and ultimately our mental health.
Subject of Research: Neural basis of uncertainty processing in decision making.
Article Title: The neural basis for uncertainty processing in hierarchical decision making.
News Publication Date: 16-Oct-2025.
Web References: http://dx.doi.org/10.1038/s41467-025-63994-y
References: Nature Communications.
Image Credits: N/A.
Keywords
Neuroscience, Psychiatric disorders, Mental health, Psychiatry, Artificial intelligence.
