UC Santa Barbara

The control of large scale neural systems promises to usher in a new era of technologies for use in treating disease, answering scientific questions, and improving human performance. Unlike other systems amenable to modern control and reinforcement learning (e.g. robots), we have many examples of high-functioning brains being used by healthy individuals. Combined with the fact that exploratory perturbations of the brain are extremely expensive relative to observation of healthy brains, this suggests that an appropriate workflow for constructing a controller should begin with inverse methods. We present a number of results showing how inverse schemes can be used to both model human behavior and neural dynamics. We show how these methods can be used to facilitate interactions between human beings and artificial agents. We conclude by suggesting that with further refinement, such schemes can be used to directly guide perturbation of the brain.