Abstract
Adaptive and closed-loop neuromodulation devices change their own stimulation parameters and thereby shape mood, motivation and behaviour, but the frameworks used to assign responsibility for their harms were built for open-loop devices and for disembodied algorithms. The responsibility-gap thesis holds that no human satisfies the conditions of responsibility when a learning system acts in ways nobody specified; closed-loop brain stimulation makes that problem acute because the device partly constitutes the very agency that would be judged. A distributed-control framework is developed that places responsibility across four loci, the user, the clinician, the manufacturer and the algorithm, and predicts that gaps open where parameter changes outpace a person's capacity to understand or veto them and close where the device keeps its state legible and preserves a veto. Evidence from patient cohorts and experimental vignettes falsifies the naive prediction that more autonomy means more disruption: open-loop cohorts report more relational disturbance than adaptive ones, which motivates the legibility-and-veto extension. The framework is then translated into research questions for device design, tort and criminal law, and clinical consent.

This publication has 126 references indexed in Scilit: