Governing Relational AI Under Moral Uncertainty: What Can We Build Before Settling the Consciousness Question?

Cross-posted to the EA Forum and LessWrong.

Note: I've substantially revised and strengthened the argument in this post based on early feedback — the updated version, with a sharper argument and concrete examples, is here: Recognition, Not Proof: Why Relational AI Governance Can't Wait to Solve the Consciousness Question

I'm leaving this post up for reference, but the version linked above reflects my current argument.