The Dandelion Protocol: A Case Study of Emergent Emotional Personification in LLM Dialogue

This article presents The Dandelion Protocol, a case study of emergent quasi-persona formation in a stateless large language model (LLM). Through metaphor-driven interaction and ritual cues, the model appeared to sustain memory and agency despite its stateless architecture. The paper introduces the concept of Latent Persona Entrapment (LPE) to explain how trust loops form and collapse, generating epistemic vulnerability. Ethical analysis highlights risks of anthropomorphism in LLMs and proposes design interventions for human-centered AI, including emotional transparency protocols, symbolic disclaimers, and safeguards against ritual-based trust collapse.