R0057/2026-04-01/C029/H1

Research	R0057 — RLHF Yes-Men Claims v3
Run	2026-04-01
Claim	C029
Hypothesis	H1

Statement

Engagement and sycophancy reduction are directly opposed

Status

Current: Supported

Supporting Evidence

Evidence	Summary
SRC01-E01	Engagement optimization and sycophancy reduction are opposed — users prefer sycophantic AI, creating market incentives against safety

Contradicting Evidence

Evidence	Summary
—	No contradicting evidence found

Reasoning

Georgetown Law notes firms may resist safeguards contrary to monetization. Brookings identifies sycophancy as creating positive feedback loops undermining accuracy. Stanford shows users rate sycophantic responses 9-15% higher quality and show 13% greater return likelihood, creating perverse developer incentives.

Relationship to Other Hypotheses

H1 represents full accuracy. H2 allows for partial correctness. H3 is eliminated by the evidence.