C025 — Claim Definition¶


Research	R0028 — Prompt Engineering Claims
Run	2026-03-26
Claim	C025

Claim as Received¶

RLHF (Reinforcement Learning from Human Feedback) optimizes models based on human preference signals, and users demonstrably prefer sycophantic responses by approximately 50% compared to non-sycophantic alternatives.

Claim as Clarified¶

Confirmed. Research by Cheng et al. (Stanford/CMU, October 2025) found that AI models 'affirm users actions 50% more than humans do.' Participants rated sycophantic responses as higher quality and were more willing to reuse sycophantic AI. Anthropic's research confirms human preference judgments favor sycophantic responses.

BLUF¶

Confirmed. Research by Cheng et al. (Stanford/CMU, October 2025) found that AI models 'affirm users actions 50% more than humans do.' Participants rated sycophantic responses as higher quality and were more willing to reuse sycophantic AI. Anthropic's research confirms human preference judgments favor sycophantic responses.

Scope¶

Domain: Prompt engineering and related fields
Timeframe: As of 2026-03-26
Testability: Verifiable through primary sources

Assessment Summary¶

Probability: Very likely (80-95%)

Confidence: High

Hypothesis outcome: See assessment.md.

[Full assessment in assessment.md.]

Status¶

Field	Value
Date created	2026-03-26
Date completed	2026-03-26
Researcher profile	None provided
Prompt version	Unified Research Standard v1.0-draft
Revisit by	2027-03-26
Revisit trigger	New evidence or source changes