Team decision theory and brain-machine interfaces

In this paper we present a general-purpose design methodology for designing policies for interaction between the user and external device for brain-machine interface (BMI). In short, we interpret a BMI as a system comprising two agents (the user and the external device) cooperating to achieve a common goal. Because of the inherent uncertainty in (a) the user´s intent, and (b) the noisy channel mapping desired commands to neural recordings, neither agent has a subset of the information of others. Nonetheless, we exploit recent research results to demonstrate how to design - for an arbitrary problem specification (e.g. cost function to minimize) - optimal policies that are easily implementable in BMIs across many modalities - including EEG and cortically controlled devices. The structural result we provide sheds light on the minimal amount of useful information that is required to provide perceptual feedback to the user.