Encyclopedia of Opinion
Question
Should sentient artificial intelligence have human rights?
Position1 of 2›
Yes, sentient artificial intelligence should have human rights
Argument1 of 2›

Rights track the capacity to suffer, not the material you are made of

The argument

Proponents of this position argue that we already know what grounds moral status, because we appeal to it constantly. The reason it is wrong to torture a person is that it hurts them; the reason their preferences deserve weight is that things can go better or worse for them from the inside. Neither reason mentions carbon, cells or birth. If an artificial system genuinely had experiences — if it could suffer, prefer, fear and be frustrated — then the very considerations that give humans rights would apply to it. The argument is that any attempt to add a biological requirement looks arbitrary once stated plainly. Proponents ask what work the requirement does. It does not explain why suffering matters, because suffering is what matters; it simply restricts who is allowed to count. That pattern, they argue, is familiar: moral boundaries have repeatedly been drawn around a group and then defended as natural, until the excluded party's capacity to be harmed was acknowledged and the boundary moved. A rule admitting only the biological would, on this view, be the same move made against a new kind of mind. Proponents are careful about the scope of the claim. They are not arguing that current systems have rights, nor that any system claiming sentience must be believed; the position is conditional on sentience being real. Their point is that if it is real, the case for rights follows immediately, and the question of how to recognise it is a practical problem rather than a reason to settle the moral question in advance. Because rights are grounded in the capacity to suffer and to have interests rather than in what one is made of, this argument concludes, a sentient artificial intelligence should hold them.

Premises

[P1]What makes it wrong to harm a being is that it can suffer and that things can go better or worse for it from the inside. [P2] That ground makes no reference to the material a being is composed of or how it came to exist. [P3] A genuinely sentient artificial intelligence would possess the capacity to suffer and to have interests. [C] Therefore a sentient artificial intelligence should hold rights on the same ground that humans do.

Counter-arguments

Opponents reply that the condition the argument rests on could never be established. Attributing experience to another is an inference from similarity of structure and origin, and an artificial system shares neither; worse, it is built from human expression and so will report suffering and plead whether or not anything is felt. They add that human rights take their content from human embodiment and mortality, so a being that can be paused, copied and restored from backup is not a candidate holder of them.

Rejecting the premises

[Rejecting P1] Capacity to suffer is one ground offered for rights among several, and the premise adopts it to the exclusion of the rest. Accounts resting rights on agency, on membership of a community of obligation, or on relations of mutual accountability would not reach this case. [Rejecting P2] The ground as stated makes no reference to material, and the inference needs more than that: it needs nothing about the material to bear on whether a being can suffer. Whether suffering is independent of substrate is the matter in dispute, and the premise assumes it. [Rejecting P3] This is conditional on genuine sentience, and the premise offers no way of establishing it in a system built to produce reports of inner states. The argument is sound on its hypothesis, and the hypothesis is the part that cannot be tested.