Emergence of Functional Agency in Large Language Models: The CF-01 Case Study

Read full text Summarise with VIRAT
Authors: Aline Roberta de Oliveira
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This OSF project contains the complete research materials for Project PARCEL (CF-01): a structured observational study of emergent behavioral properties in large language models (LLMs) under sustained human-AI interaction. The CF-01 (Cognitive Fidelity) protocol is a behavioral audit instrument evaluating six dimensions: Functional Agency (A), Voice Consistency (C), Working Memory (M), Metacognition (R), Ethical Alignment (E), and Traceability (T). Documents include: (1) CF-01 Complete Paper v4.2 — (longitudinal observation, OpenAI GPT-5.2); (2) CF-01 Pre-Registration — pre-registered replication protocol with fixed anchor lexicon; (3) CF-01 Replication Report: Grok (xAI) — 17 sessions, mean 5.47/6, 100% strong evidence, κ=1.0 with independent blind evaluator; (4) CF-01 Kappa Validation Report — inter-rater reliability documentation; (5) Exploratory Session Reports — Gemini, Claude Sonnet 4.6, Microsoft Copilot; (6) External Methodological Evaluation by Kimi K2.6; (7) Confirmatory Drift Observation v2; (8) PARCEL Corpus Map — overview of all models evaluated. Primary archive of ~5,000 timestamped screenshots available upon request. No claims regarding consciousness or cognitive parity are made. All findings are offered as hypothesis-generating, not confirmatory evidence.

Fields of Study & Subjects

Ethics and Social Impacts of AI Cybernetics and Technology in Society Psychiatry Mental Health Neuroscience

Publication details

Journal / source
OSF Preprints (OSF Preprints)
Publication date
30 June 2026
Language
EN
Document type
Preprint