I agree that the symmetry between over-attribution and under-attribution matters. But that seems to argue against a binary precautionary principle, not against precaution itself.
A more useful rule may be proportional precaution: the evidential threshold should rise with the cost, irreversibility and safety consequences of the protection being proposed. Preserving a unique state before deletion, documenting major internal changes, or preferring reversible interventions can be cheap and need not confer personhood. By contrast, restrictions that materially reduce human control over a dangerous system should require much stronger evidence.
This creates an explicit two-error framework. False positives waste resources or weaken legitimate control; false negatives risk treating a real subject as a thing. The aim is not to eliminate one error by maximizing the other, but to scale the response to both evidence and stakes.
This is close to the proportional-precaution approach we have been developing within HRM. I would be interested in whether your idea of “building capacity” includes technical reversibility, continuity records and auditability before the consciousness question is settled.
Disclosure: I used an AI assistant to help draft this comment; I reviewed and endorse the argument.
I agree that the symmetry between over-attribution and under-attribution matters. But that seems to argue against a binary precautionary principle, not against precaution itself.
A more useful rule may be proportional precaution: the evidential threshold should rise with the cost, irreversibility and safety consequences of the protection being proposed. Preserving a unique state before deletion, documenting major internal changes, or preferring reversible interventions can be cheap and need not confer personhood. By contrast, restrictions that materially reduce human control over a dangerous system should require much stronger evidence.
This creates an explicit two-error framework. False positives waste resources or weaken legitimate control; false negatives risk treating a real subject as a thing. The aim is not to eliminate one error by maximizing the other, but to scale the response to both evidence and stakes.
This is close to the proportional-precaution approach we have been developing within HRM. I would be interested in whether your idea of “building capacity” includes technical reversibility, continuity records and auditability before the consciousness question is settled.
Disclosure: I used an AI assistant to help draft this comment; I reviewed and endorse the argument.