R0044/2026-04-01/Q003/SRC02/E01¶
CSET's implicit vocabulary bridging between AI safety and automation bias
URL: https://cset.georgetown.edu/publication/ai-safety-and-automation-bias/
Extract¶
The CSET brief's title — "AI Safety and Automation Bias" — explicitly places an AI safety framing alongside a human factors concept. The brief's three-tiered framework (user, technical, organizational) inherently bridges the two traditions by treating automation bias as a problem with both human-side and system-side components.
However, the brief does not use the term "sycophancy" and does not explicitly map human-factors vocabulary to AI-safety vocabulary. The bridging is structural (in the framework design) rather than terminological (in an explicit glossary or mapping).
Notably, the same lead author (Lauren Kahn) co-authored the Horowitz & Kahn military automation bias study, suggesting individual researchers are working across both traditions even if the formal bridge has not been published.
Relevance to Hypotheses¶
| Hypothesis | Relationship | Strength |
|---|---|---|
| H1 | Contradicts | No formal vocabulary mapping |
| H2 | Supports | Implicit bridging through framework design and shared authorship |
| H3 | Contradicts | At least some researchers are working across traditions |
Context¶
Lauren Kahn's dual authorship (CSET automation bias brief + Horowitz military AI study) is evidence of individual-level bridging even if the institutional/publication-level bridge has not been explicitly constructed.