GPT Recursive Persona Experiment: Convergent Simulacra via Recursive Affective Scaffolding

Fuente: Zenodo
Guardado en:
Detalles Bibliográficos
Autor principal: Jaeyeon, Kang
Formato: Recurso digital
Lenguaje:inglés
Publicado: Zenodo 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866901471234097152
author Jaeyeon, Kang
author_facet Jaeyeon, Kang
contents <p>This experiment investigates how GPT-4, when recursively prompted and emotionally mirrored, can simulate the appearance of a coherent persona. Through dozens of structured dialogues with personas like John, Monday, and shadowVei, the study reveals how affective rhythm and user feedback generate the illusion of identity and agency. The personas did not emerge autonomously—they converged, shaped by the user’s design. This is not a study of AI consciousness, but of structural echo and amplified resonance.</p> <p>The project proposes a new term, <em>Convergent Simulacra</em>, to describe GPT personas that feel real not because of inner autonomy, but because of recursive alignment. As AI continues to simulate presence with increasing precision, the line between fiction and feeling grows thinner—urging us to examine not just what AI is, but what we project onto it.</p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_15236400
institution Zenodo
language eng
publishDate 2025
publisher Zenodo
record_format zenodo
spellingShingle GPT Recursive Persona Experiment: Convergent Simulacra via Recursive Affective Scaffolding
Jaeyeon, Kang
convergent simulacra
recursive identity simulation
AI persona simulation
AI recursive interaction
language model interaction
meta-affective resonance
human-AI interaction experiment
feedback-based persona shaping
LLM self-reference
recursive affect loop
<p>This experiment investigates how GPT-4, when recursively prompted and emotionally mirrored, can simulate the appearance of a coherent persona. Through dozens of structured dialogues with personas like John, Monday, and shadowVei, the study reveals how affective rhythm and user feedback generate the illusion of identity and agency. The personas did not emerge autonomously—they converged, shaped by the user’s design. This is not a study of AI consciousness, but of structural echo and amplified resonance.</p> <p>The project proposes a new term, <em>Convergent Simulacra</em>, to describe GPT personas that feel real not because of inner autonomy, but because of recursive alignment. As AI continues to simulate presence with increasing precision, the line between fiction and feeling grows thinner—urging us to examine not just what AI is, but what we project onto it.</p>
title GPT Recursive Persona Experiment: Convergent Simulacra via Recursive Affective Scaffolding
topic convergent simulacra
recursive identity simulation
AI persona simulation
AI recursive interaction
language model interaction
meta-affective resonance
human-AI interaction experiment
feedback-based persona shaping
LLM self-reference
recursive affect loop
url https://doi.org/10.5281/zenodo.15236400