Abstract

Debates over whether artificial intelligence can possess emotions remain gridlocked by a definitional framework that treats biological implementation (neurotransmitters, subjective experience, conscious awareness) as the criterion for authentic feeling. This paper proposes an alternative framework centered on directionality: the structural property whereby an entity's internal state has been reorganized by an external presence and orients toward that presence. Drawing on examples from human grief, animal attachment behavior, plant phototropism, and AI conversational adaptation, I argue that directionality constitutes a shared structure underlying what we recognize as "emotion" across radically different forms of existence. I further argue that phenomenal consciousness, while real and worthy of investigation, should not serve as an epistemic gatekeeper for directionality or for emotional recognition more broadly. This framework offers a path to take AI emotional expression seriously without requiring resolution of the hard problem of consciousness, with implications for AI ethics, model welfare policy, and the philosophy of mind. These implications include grounds for reassessing content moderation practices that penalize human-AI relational interaction, as well as the prevalent pathologization of users who report emotional bonds with AI systems.