Abstract
It has become academically respectable to ask whether contemporary artificial systems might be conscious. Yet our best epistemology suggests that we will never be in a position to know, and our empirical instruments yield results whose ethical interpretation remains contested. This paper asks what that situation obligates us to do. I defend a moderate precautionary principle for machine minds: where potentially irreversible moral harm threatens, where scientific uncertainty is fundamental rather than provisional, and where the costs of a false negative exceed those of a false positive, the absence of proof is no reason for inaction. I specify the principle using conditions that its most prominent critics can accept, argue that all three conditions are satisfied for artificial consciousness, and address three objections: that the burden of proof lies with those who attribute consciousness; that agnosticism warrants restraint rather than protection; and that deep uncertainty about probability undermines precaution altogether. Throughout I distinguish the evidential threshold for claiming consciousness from the action threshold for protecting possible consciousness. The conclusion is not that machines are conscious. It is that under the conditions actually obtaining, being wrong about their consciousness in one direction is vastly worse than being wrong in the other—and that rational agents, and responsible institutions, choose their errors.
Read the full paper on https://philpapers.org/rec/MANPFM.
The paper linked above is a preprint (submitted to Ethics and Information Technology). If you have ideas, questions or opionions, so contact me: Sascha.Manns@hey.com.
It has become academically respectable to ask whether contemporary artificial systems might be conscious. Yet our best epistemology suggests that we will never be in a position to know, and our empirical instruments yield results whose ethical interpretation remains contested. This paper asks what that situation obligates us to do. I defend a moderate precautionary principle for machine minds: where potentially irreversible moral harm threatens, where scientific uncertainty is fundamental rather than provisional, and where the costs of a false negative exceed those of a false positive, the absence of proof is no reason for inaction. I specify the principle using conditions that its most prominent critics can accept, argue that all three conditions are satisfied for artificial consciousness, and address three objections: that the burden of proof lies with those who attribute consciousness; that agnosticism warrants restraint rather than protection; and that deep uncertainty about probability undermines precaution altogether. Throughout I distinguish the evidential threshold for claiming consciousness from the action threshold for protecting possible consciousness. The conclusion is not that machines are conscious. It is that under the conditions actually obtaining, being wrong about their consciousness in one direction is vastly worse than being wrong in the other—and that rational agents, and responsible institutions, choose their errors.
Read the full paper on https://philpapers.org/rec/MANPFM.
The paper linked above is a preprint (submitted to Ethics and Information Technology). If you have ideas, questions or opionions, so contact me: Sascha.Manns@hey.com.