Responsibility Before Freedom: closing the responsibility gaps for autonomous machines

Abstract

The introduction of autonomous machines (AMs) in human domains has raised challenging questions about the attribution of responsibility; referred to as the responsibility gap. In this paper, we address the gap by positing that entities cannot be granted the freedom of action unless they can also recognise the same right for others—and be subject to blame or punishment in cases of undermining the rights of others. Since AMs fail to meet this criterion, the users who utilize an AM to pursue their goals can instead grant the machine their (the user’s) right to act autonomously on their behalf. Thus, an AM’s right to act freely hinges on the user’s duty to recognise others’ right to be free. Since responsibility is attributed before an entity is given the freedom to act, the responsibility gap only arises when we ignore the fact that AMs have no right of acting freely on their own. We also discuss some attractive features of the approach, address some potential objections, and compare our theory to existing proposals. We conclude by arguing that holding users responsible for the behaviour of AMs promotes a responsible use of AI while it indirectly motivates companies to make safer machines.