Hybrid Intention and Augmented Agency: Allocation of Criminal Culpability in Neuroprosthetic Control

Abstract

Neuroprosthetic interfaces are entering clinical settings and reshaping what it means to control one’s own actions. Criminal courts may soon face harmful outcomes linked to devices that decode neural or muscular signals and translate them into physical action. In such cases, criminal law risks either placing guilt too quickly on the nearest human actor or diffusing responsibility until no one can be held to account. This paper develops a framework that keeps culpability attached to human and organisational actors while accounting for technologically mediated agency. Hybrid intention describes cases in which a human purpose remains identifiable but the system materially shapes how that purpose is translated into conduct. Attribution must therefore rest on reconstructable evidence rather than formal human presence. The paper introduces the X-V-L gate, requiring case-level explainability, prior validation, and sufficiently complete logging before a system-mediated output can support attribution. It then combines this threshold with a control-window approach that locates usable control across design, configuration, operation, and emergency. Two neuroprosthetic scenarios show how differences in warning, interruption, and system records can alter the attribution analysis. The result is a court-usable decision rule that separates failure of attribution to one actor from any independent case against another.