Skip to content

R0044/2026-03-29/Q004 — Query Definition
Research R0044 — Expanded Vocabulary Research
Run 2026-03-29
Query Q004

Query as Received

The DoD CaTE (Calibrated AI Trust and Expectations) center was identified as having the most sophisticated regulated-industry vocabulary for this problem. What has CaTE published about calibrating trust in AI systems, and does their work address the system-side behavior (AI adjusting output to match user expectations) or only the human-side behavior (users trusting AI too much)?

Query as Clarified

  • Subject: Publications and outputs of the DoD Center for Calibrated Trust Measurement and Evaluation (CaTE), operated by CMU SEI
  • Scope: The scope of CaTE's published work, specifically whether it addresses system-side behavior (AI adjusting output) vs. only human-side behavior (users trusting AI inappropriately)
  • Evidence basis: Published documents, guidebooks, research papers, and organizational descriptions from CaTE/SEI
  • Embedded assumption: The query states CaTE has "the most sophisticated regulated-industry vocabulary" — this is taken as a research input to be tested, not as an axiom

Ambiguities Identified

  1. "CaTE" name: The query uses "Calibrated AI Trust and Expectations" but the actual name appears to be "Center for Calibrated Trust Measurement and Evaluation." This difference is minor but noted.
  2. "System-side behavior" in the CaTE context could mean: (a) the AI system adjusting its output to match user expectations (sycophancy), or (b) the AI system being designed with properties that enable calibrated trust (transparency, explainability). CaTE appears to focus on (b).

Sub-Questions

  1. What has CaTE published since its 2023 launch?
  2. Does CaTE's framework address AI system output behavior, or only system properties (transparency, explainability, reliability) that enable human trust calibration?
  3. Does CaTE use the term "sycophancy" or otherwise address the phenomenon of AI systems reinforcing user expectations?
  4. Is CaTE's vocabulary truly the most sophisticated in regulated industries for this problem?

Hypotheses

ID Hypothesis Description
H1 CaTE addresses both system-side and human-side behavior CaTE's published work examines both how AI systems should behave to enable calibrated trust AND how humans should calibrate their trust
H2 CaTE addresses only human-side behavior CaTE's work focuses exclusively on measuring and calibrating human trust in AI systems, without addressing how AI systems should adjust their own behavior
H3 CaTE addresses system properties (design) but not system output behavior CaTE examines AI system design properties (reliability, transparency, explainability) that enable trust calibration, but does not address the system actively adjusting its output to match or counteract user expectations