Skip to content

R0043/2026-03-28/Q001

Query: What terms do different industries and disciplines use to describe AI behavior that prioritizes user agreement, comfort, or satisfaction over accuracy, correctness, or safety? Map the complete vocabulary across: AI safety research, defense/military AI, healthcare AI, financial services AI, aviation/FAA, academic integrity, enterprise software evaluation, and UX/product design.

BLUF: The vocabulary is not missing — it is systematically asymmetric. Regulated industries have mature terminology for the human side (automation bias, complacency, overtrust) while AI safety alone has terminology for the system side (sycophancy). This human-side/system-side divide reflects when each domain's vocabulary was developed: traditional automation era vs. adaptive AI era. No shared bridging vocabulary exists.

Answer: H3 (Partial vocabulary with systematic gaps) · Confidence: High


Summary

Entity Description
Query Definition Question as received, clarified, ambiguities, sub-questions
Assessment Full analytical product
ACH Matrix Evidence × hypotheses diagnosticity analysis
Self-Audit ROBIS-adapted 4-domain process audit

Hypotheses

ID Statement Status
H1 Rich cross-domain vocabulary exists across all domains Partially supported
H2 No cross-domain vocabulary exists outside AI safety Eliminated
H3 Partial vocabulary with systematic human-side/system-side gaps Supported

Cross-Domain Vocabulary Map

Domain Primary Terms Framing Specificity
AI Safety Research Sycophancy, reward hacking, regressive/progressive sycophancy, social/propositional sycophancy System-side (what the AI does) High — sub-taxonomy with measurement terms
Defense/Military Calibrated trust, overtrust, distrust, appropriate trust, automation bias Human-side (what the operator does) High — institutionalized with CaTE center
Healthcare Acquiescence problem, automation bias, alert fatigue, deskilling, commission/omission errors Mixed — acquiescence is partially system-side High — measurable patient safety impact
Aviation/FAA Automation complacency, over-trust, loss of situational awareness Human-side Medium — experts acknowledge terms are insufficient for AI
EU Regulation Automation bias (Article 14 EU AI Act) Human-side (legal obligation on deployers) Medium — single legal term
U.S. Standards Overreliance, automation bias, inappropriate anthropomorphizing, emotional entanglement Human-side Medium — multiple terms but no system-side
Financial Services Model risk management, model validation Generic — no sycophancy-specific terms Low — borrows from general AI governance
Academic Integrity Borrows "sycophancy" from AI safety; "grade inflation" as downstream effect Borrowed Low — no domain-native terminology
Enterprise Evaluation Agreeableness bias (in LLM evaluators), hallucination rate Mixed Low — emerging vocabulary
UX/Product Design People-pleasing, yes-man, confirmation bias amplification Informal system-side Low — colloquial, not formalized

Searches

ID Target Type Outcome
S01 AI safety sycophancy definitions WebSearch 4 selected / 6 rejected — rich AI safety vocabulary confirmed
S02 Cross-domain automation bias terminology WebSearch 3 selected / 7 rejected — automation bias as cross-domain umbrella
S03 UX/product design terms WebSearch 2 selected / 8 rejected — informal vocabulary only
S04 Defense and aviation terminology WebSearch 4 selected / 26 rejected — richest regulated-industry vocabulary
S05 Financial services AI terminology WebSearch 1 selected / 9 rejected — generic terms only
S06 Academic integrity AI terms WebSearch 2 selected / 8 rejected — borrows from AI safety

Sources

Source Description Reliability Relevance Evidence
SRC01 Parasuraman & Manzey 2010 High High 1 extract
SRC02 NN/g Sycophancy High High 1 extract
SRC03 TechPolicy.Press Medium-High High 1 extract
SRC04 CSET Automation Bias High High 1 extract
SRC05 EU AI Act Article 14 High High 1 extract
SRC06 NIST AI RMF High High 1 extract
SRC07 DoD CaTE High High 1 extract
SRC08 Aviation HFACS Medium-High High 1 extract
SRC09 Healthcare Acquiescence Medium-High High 1 extract
SRC10 Vocabulary Gap paper Medium High 1 extract

Revisit Triggers

  • Publication of updated HFACS taxonomy incorporating AI-specific terms
  • NIST or EU AI Act amendments introducing system-behavior terminology
  • Financial services regulators issuing AI-specific guidance that names sycophancy-adjacent phenomena
  • Publication of a cross-domain vocabulary standard or ontology