Type: conference-paper
Abstract: Large Language Models are increasingly used to simulate identities with distinct personality traits, communication styles, and behavioural patterns across applications such as conversational agents and synthetic user simulation. However, LLM-based personas are commonly specified through ad hoc prompts or biographical sketches, leaving room for ambiguity, inconsistent interpretation, and limited portability. Existing persona ontologies and structured user models only partially address these limitations, as they are not mainly designed to support persona enactment by LLMs. We introduce the LLM Persona ontology, structured following the OWL formalism around two foundational pillars: Identity and Behaviour. We also propose a downstream evaluation using PersonaGym to assess whether ontology-based representations can support persona enactment, as a first step toward formalizing personas for LLM-based personalization.
Source: PORTO@Iris
Institutions: Politecnico di Torino, EURECOM
Language: English
Subfield: Human-Computer Interaction
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7216741585