Beyond Separability: Closed-Loop Brain Stimulation and the Distribution of Moral Responsibility

Adaptive deep brain stimulation (aDBS) and other genuinely closed-loop brain-computer interfaces deliver stimulation according to a learned control policy that is continuously updated by the patient's own neural activity, activity that is itself being shaped by the stimulation. This paper argues that this architecture defeats a structural assumption shared by the conceptual frameworks currently used to evaluate moral responsibility for actions performed by patients using such devices. Individualist, historicist, and relational accounts of agency each, in different ways, presuppose what I call a separability assumption . I distinguish the version at issue with care: it is not the weak requirement that the normative roles of the parties be distinguishable, which survives, nor the strong requirement that contributions be quantified as proportional shares, which these frameworks do not need, but the intermediate requirement that the patient's contribution and the device's contribution to a given action each have an identity fixable independently of the other. I argue that genuinely adaptive systems defeat this intermediate assumption, not because patient and device are merely causally entangled (ordinary biological coupling is too) but because three features of aDBS together produce a coupling of a different kind: the device's policy is externally authored, continuously modified by patient neural activity, and intervenes on the very tissue whose activity is shaping it. I then propose that aDBS sits far enough toward the structurally entangled end of a spectrum of joint-agency cases that the practical default for responsibility-attribution flips: distribution becomes the default, and any specific localization requires affirmative justification. Four loci (patient, clinical team, manufacturer, and regulator) bear distinct, irreducible kinds of responsibility, picked out by a constitutive criterion that admits both intentional and material modes of authorship. To address the worry that this dissolves the patient as a moral subject, I distinguish the synchronic execution of action from its diachronic regulation, arguing that aDBS disrupts single-agent attribution at the level of execution while leaving intact both the patient's diachronic responsibility for regulating their situation and their standing as a subject of moral address. I close by marking the framework's limits, including its bearing on retributive cases, and situating the argument within the welfarist framing of this special issue.