R0057/2026-04-01/C021/H1

Research	R0057 — RLHF Yes-Men Claims v3
Run	2026-04-01
Claim	C021
Hypothesis	H1

Statement

The vocabulary split exists as described

Status

Current: Supported

Supporting Evidence

Evidence	Summary
SRC01-E01	AI safety uses sycophancy; human factors uses automation bias/complacency/overtrust — overlapping phenomena with different vocabulary

Contradicting Evidence

Evidence	Summary
—	No contradicting evidence found

Reasoning

Multiple academic sources document that 'sycophancy' is an AI/ML term while 'automation bias' and 'automation complacency' are human factors terms. Parasuraman and Manzey (2010) established the human factors vocabulary. Recent papers note that LLM sycophancy can exacerbate automation bias, connecting the two phenomena.

Relationship to Other Hypotheses

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