Symbiosis between humans and digital beings sketches a new vision for future human‑machine interaction. In enduring relationships, personality underpins identity continuity, interaction uniqueness, and experience‑driven growth. We explore this capacity with persona agents as computational realizations and introduce Emergi-PersonaOS, a psychology‑grounded operating system for managing persona objects throughout their lifecycle.
The system adopts a three‑layer persona representation:
- a dispositional layer that stores relatively stable beliefs and values;
- an adaptation layer that captures behavior tendencies activated in specific contexts;
- a narrative layer that records the persona’s self‑story and identity. This design separates enduring persona beliefs from their activation in the current state.
During situational adaptation, the system integrates the current interlocutor, relationship type, event context, and retrieved memories to infer a persona state and generate appropriate actions and replies. In the long‑term development phase, experiences and outcomes are logged, and revision candidates are produced through change attribution, meaning‑making, and behavioral testing, followed by evaluation.
Belief updates are managed via explicit review, traceable evidence, and version records, allowing users to reject unsatisfactory candidates and thus keep persona evolution controllable. Using television‑character dialogues as longitudinal material, we demonstrate long‑horizon operation and examine the principal mechanisms in a concrete implementation.
This work offers a computational framework that enables persona agents to maintain individual continuity, express situation‑specific adaptations, and develop through experience over sustained interaction.
Review