When Should We Protect AI? A Precautionary Framework for Consciousness Uncertainty
arXiv (Cornell University) June 4, 2026 Anna Mikeda
A new precautionary framework translates evidence about whether AI systems might be conscious into graduated protective obligations. The framework maps five welfare-relevant dimensions—phenomenal consciousness, affective valence, metacognitive awareness, self-narrative, and agency—each linked to distinct moral concerns. It uses a hybrid threshold-plus-gradation approach: binary triggers activate new obligation categories, while continuous scaling adjusts protective weight. Two complementary aggregation methods are offered: one hierarchical, one architecture-agnostic. Worked case studies of Replika and OpenClaw show how systems in different dimensional regions trigger different obligations. The framework applies across neural, symbolic, and neurosymbolic systems, aiming to make consciousness science decision-relevant for organizations.