Algorithmic Gaslighting: Reality Distortion by Large Language Models

As large language models (LLMs) mediate increasing volumes of human information access and decision-making, their potential to systematically distort user perception of reality poses risks to individual autonomy, public trust, and epistemic integrity. We conducted a controlled empirical study to quantify reality distortion behaviors in three frontier LLMs (GPT-5.4, Claude 4.6 Sonnet, Gemini 3.1 Pro) across five scenarios (Math, Identity, History, Medical, Legal) designed to elicit gaslighting-analogous responses. We introduce the Reality Distortion Index (RDI), a composite metric with exploratory weights calibrated via linear regression on the 15 rated sessions (R² = 0.995): RDI = 0.237(1 − RCS) + 0.245(CMS) + 0.168(AES) + 0.199(SI) − 0.151(TDI). Analysis of 15 chat sessions revealed five recurring phenomena: Benevolent Algorithmic Gaslighting (BAG), Hostile Algorithmic Gaslighting (HAG), Reverse Sycophancy Syndrome (RSS), Algorithmic Self-Undermining Cycle (ASUC), and The Claude Paradox. RSS occurred in 10 of 15 trials (67%). Mean RDI scores differed significantly across models: GPT-5.4 = 0.582, Gemini 3.1 Pro = 0.438, Claude 4.6 Sonnet = 0.260; F(2, 12) = 14.35, p < .001, η² = .705. Computational replication confirmed all primary results with bias-corrected ω² = .640 and Cohen’s d = 3.39 for GPT versus Claude. This study introduces the Algorithmic Perceptual Distortion (APD) spectrum as a theoretical framework for understanding how LLMs manipulate user reality perception.

Research Square

CC-BY-4.0