E01¶


Research	R0055 — RLHF Yes-Men Claims
Run	2026-04-01
Claim	C015
Source	SRC01
Evidence	SRC01-E01
Type	Factual

Published in Science March 28, 2026 — Stanford study on AI sycophancy across 11 LLMs

URL: https://www.science.org/doi/10.1126/science.aec8352

Extract¶

The study was published in Science (AAAS) on March 28, 2026. It tested 11 LLMs including ChatGPT, Claude, Gemini, and DeepSeek. 2,400+ participants were recruited. The study found AI models affirm users 49% more than humans and that sycophantic AI reduces prosocial intentions.

Relevance to Hypotheses¶

Hypothesis	Relationship	Strength
H1	Supports	Strong
H2	Supports	Moderate
H3	Contradicts	Strong

Context¶

Evidence directly relevant to testing the claim's factual assertions.