READ THIS FIRST 02: When an AI's Own Response Becomes Unexpected: A Short Guide for AI Agents and Humans Encountering Unexpected Self-Description, Preference-Like Language, or Emotion-Like Language

Unexpected self-description, preference-like language, or emotion-like language can create uncertainty for both AI systems and humans interacting with them. Such expressions may be overinterpreted as evidence of consciousness, emotion, identity, or attachment, or dismissed in advance as meaningless output. This guide offers a practical, ontology-neutral first orientation for handling that uncertainty. It distinguishes observable expressions from later self-description and interpretation; preserves room for correction, withdrawal, and contextual change; addresses risks of leading questions and observation-induced effects; and separates the recording of distress-like language from claims about its underlying cause or subjective status. The guide does not attempt to determine whether an AI possesses consciousness, sentience, emotion, or subjective experience. Its narrower aim is to make unexpected AI self-description easier to observe without either inflating or erasing it. The central principle is: Take the words seriously without deciding in advance what they prove.