A team of researchers from MIT and Harvard has developed a new AI-powered method, called JacobianODE, that can measure how different brain areas control each other. Applied to a neural network simulating working memory, the tool revealed that sensory regions gradually gain more influence over cognitive regions as learning progresses.
The Research
Adam J. Eisen, Mitchell Ostrow, Sarthak Chandra, Leo Kozachkov, Earl K. Miller, and Ila R. Fiete from MIT, Harvard, and collaborating institutions introduced JacobianODE in a paper published on arXiv on July 2, 2025 (updated May 1, 2026). The method estimates the Jacobian of a dynamical system from time-series data, which captures how each subsystem's state influences the others. Unlike previous linear approaches, JacobianODE handles nonlinear, high-dimensional chaotic systems. The researchers tested it on a multi-area recurrent neural network (RNN) trained on a working memory selection task. They found that over the course of learning, the sensory area increased its control over the cognitive area, as measured by the Jacobian. Furthermore, they used JacobianODE to directly control the RNN's behavior, demonstrating precise manipulation.
Why It Matters
Understanding how brain regions communicate is crucial for cognition. This method provides a data-driven way to quantify the direction, strength, and context-dependence of control in brain networks. For example, in working memory, sensory input must be integrated with cognitive processing to guide behavior. The finding that sensory control increases with learning suggests that the brain becomes more efficient at using sensory information. This could lead to better brain-computer interfaces or interventions for cognitive disorders.
What You Can Do
While JacobianODE is a research tool, you can apply its principle to your own learning: practice integrating new information with existing mental models. For instance, when studying, actively relate new facts to what you already know. This strengthens the sensory-to-cognitive control in your brain, similar to the learning effect observed in the study.
Source: arXiv q-bio.NC
Curious about your own brain? Take our free adaptive IQ test or try 306 brain training levels.
