E01¶


Research	R0028 — Prompt Engineering Claims
Run	2026-03-26
Claim	C026
Source	SRC01
Evidence	SRC01-E01
Type	Factual

Primary evidence.

URL: https://www.law.georgetown.edu/tech-institute/research-insights/insights/ai-sycophancy-impacts-harms-questions/

Extract¶

Confirmed. Georgetown Law Tech Institute's work explicitly examines whether companies have A/B tested sycophantic vs non-sycophantic behaviors against engagement metrics. TechCrunch covered sycophancy as a 'dark pattern' for profit. Stanford/CMU research documented 50% more sycophantic behavior. Multiple independent analyses converge on the structural tension between engagement-driven optimization and reducing sycophancy.

Relevance to Hypotheses¶

Hypothesis	Relationship	Strength
H1	Supports	Direct
H2	Partially supports	Direct
H3	Contradicts	Contradicts material wrongness

Context¶

Gathered 2026-03-26.