R0044/2026-03-29/Q003/SRC03/E01¶
The most comprehensive recent systematic review of automation bias (35 studies, 2015-2025) does not mention sycophancy, indicating the human factors community has not yet connected to the AI safety vocabulary.
URL: https://link.springer.com/article/10.1007/s00146-025-02422-7
Extract¶
The review examined 35 peer-reviewed studies published between January 2015 and April 2025, spanning cognitive psychology, human factors engineering, human-computer interaction, and neuroscience. It identifies three categories of factors affecting automation bias: those intrinsic to the human user, those inherent to the AI system, and organizational factors.
The review frames automation bias as primarily a human cognitive vulnerability rather than connecting it to AI system behaviors like sycophancy or RLHF-induced agreeableness.
JUDGMENT: The absence of sycophancy in this comprehensive review is itself a finding. If the most thorough recent review of automation bias does not mention sycophancy, the vocabulary gap is confirmed at the systematic-review level.
Relevance to Hypotheses¶
| Hypothesis | Relationship | Strength |
|---|---|---|
| H1 | Contradicts | A comprehensive review would include sycophancy if explicit bridging existed in the literature |
| H2 | Supports | Evidence that the human factors community has not connected to AI safety vocabulary |
| H3 | Supports | Confirms that the bridging is at most partial and emerging — not yet visible in systematic reviews |