Request Tone as a Behavioral Modifier in Large Language Models

OSF Preprints (OSF Preprints) 2026 Preprint Open access

Read full text
Authors: Muhammed Mussaab Ibrahem
DOI: 10.17605/osf.io/geyrk
Cited by: 0

Abstract

Pre-registration for a large-scale empirical study examining whether request tone (Supportive / Neutral / Pressure) measurably affects LLM generative behavior across six indicators: reasoning depth, factual accuracy, hallucination, logical coherence, clarity, and response stability. 6,000 responses · 4 models · Hybrid evaluation · H1–H4 confirmatory

Fields of Study & Subjects

Speech and dialogue systems Ethics and Social Impacts of AI Natural Language Processing Techniques

Publication details

Journal / source
OSF Preprints (OSF Preprints)
Publication date
1 January 2026
Language
EN
Document type
Preprint