Skip to content

R0043/2026-04-01/Q003 — Query Definition

Query as Received

Has the vocabulary gap itself been identified as a problem in the AI safety or AI governance literature? Are there researchers or organizations working to create a shared taxonomy that bridges AI safety terminology with regulated-industry terminology?

Query as Clarified

This is a meta-question: has the gap documented in Q001 and Q002 been recognized by others? Two sub-questions: (1) Has the terminology fragmentation been identified as a problem? (2) Are there active efforts to solve it?

Embedded assumption: that the vocabulary gap is a "problem" worth solving. This should be tested — some researchers may argue that domain-specific terminology is appropriate and a shared taxonomy is unnecessary or even counterproductive.

BLUF

The AI terminology gap has been identified as a problem by multiple researchers and organizations, though not always in the specific framing of "sycophancy vocabulary." Trilateral Research (2025) explicitly diagnosed the gap and proposed solutions including minimal glossaries and translation tables. The CSIRO/UNSW team proposed a harmonised terminology framework for AI evaluation. The Roytburg & Miller (2025) network analysis of 6,442 papers found 83% homophily between AI safety and ethics communities, with only 1% of authors bridging the divide. Multiple glossary efforts exist (IAPP, Gov4AI, NIST, Cyber Risk Institute), but none specifically addresses sycophancy. The gap between AI safety and regulated-industry terminology is a recognized subset of the broader AI governance terminology problem.

Scope

  • Domain: AI safety and AI governance literature
  • Timeframe: Current as of April 2026, with focus on 2024-2026 developments
  • Testability: Verified by searching for published research, reports, and organizational initiatives addressing the terminology gap

Assessment Summary

Probability: N/A (open-ended query)

Confidence: Medium-High

Hypothesis outcome: H1 (gap recognized) is supported. H2 (gap not recognized) is eliminated. H3 (partially recognized as subset of broader problem) is the most accurate characterization.

[Full assessment in assessment.md.]

Status

Field Value
Date created 2026-04-01
Date completed 2026-04-01
Researcher profile Phillip Moore
Prompt version Unified Research Methodology v1
Revisit by 2027-01-01
Revisit trigger NIST or ISO publishes cross-domain AI terminology standard; major conference session on AI terminology unification; regulatory body adopts "sycophancy" or equivalent