The Edge of Sentience: Risk and Precaution in Humans, Other Animals and AI

The precautionary framework for moral consideration under uncertainty. Part V ('Preparing for Artificial Sentience') is essential: the Gaming Problem (LLMs can game sentience criteria because their training data contains information about what convincing evidence looks like), deep computational markers over behavioral tests, and the Run-Ahead Principle (regulate for future trajectories, not just current risk). Birch's key insight: behavioral evidence from LLMs is systematically unreliable — you have to look at the computation, not the output. Introduces the 'zone of reasonable disagreement' where precaution should apply.