Flight Simulator for the Brain Reveals How We Learn and Why Minds Go Off Course (now.tufts.edu)

Researchers introduced CogLinks, a biologically grounded computational model that simulates how neural circuits assign value and adapt under uncertainty. Unlike opaque AI “black boxes,” CogLinks models neuron-like units and realistic connectivity, letting scientists trace how virtual neurons map structure to function. In silico experiments showed that weakening the connection between the prefrontal cortex (PFC) and the mediodorsal thalamus pushed the system toward slower, habit-driven learning. The team validated those predictions with an fMRI study where human volunteers played a task with hidden rule changes: the PFC handled planning, the striatum tracked habitual responses, and the mediodorsal thalamus activated when subjects detected rule shifts and reconfigured strategy—acting as a switchboard between flexible and habitual systems. The work, published in Nature Communications, is significant because it bridges single-cell biology, brain imaging, and behavior to reveal mechanistic pathways that can misfire in psychiatric disorders (e.g., schizophrenia, OCD, ADHD). CogLinks offers an interpretable “flight simulator for the brain” that can test how circuit-level changes or disease-linked molecular perturbations disrupt inference and flexibility, enabling hypothesis-driven biomarkers and more precise therapeutic targets. The approach paves the way for “algorithmic psychiatry,” where causal computational models help link genetics and neurobiology to cognitive symptoms and treatment strategies.