Skip to content

R0044/2026-04-01/Q004

Query: 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)?

BLUF: CaTE has published one primary guidebook (TEVV of LAWS, April 2025). Its scope covers both system trustworthiness evaluation and operator trust measurement, but emphasizes the human side. CaTE does not address AI systems adjusting output to match user expectations, does not use sycophancy vocabulary, and does not constrain AI output behavior. Its "calibrated trust" concept is a human-side calibration: matching operator trust to system capability.

Probability: N/A (open-ended query) | Confidence: Medium

Correction: The query refers to "Calibrated AI Trust and Expectations" but CaTE actually stands for "Center for Calibrated Trust Measurement and Evaluation."


Summary

Entity Description
Query Definition Query text, scope, status
Assessment Full analytical product with reasoning chain
ACH Matrix Evidence x hypotheses diagnosticity analysis
Self-Audit ROBIS-adapted 5-domain audit (process + source verification)

Hypotheses

ID Hypothesis Status
H1 CaTE addresses AI output behavior Eliminated
H2 Both sides, human emphasis Supported
H3 Human-side only Eliminated

Searches

ID Target Results Selected
S01 CaTE center overview 10 2
S02 CaTE guidebook 10 1

Sources

Source Description Reliability Relevance
SRC01 CaTE TEVV Guidebook High High
SRC02 SEI Annual Review High Medium-High
SRC03 DefenseScoop launch Medium-High Medium-High

Key Insight

CaTE's "calibrated trust" concept answers the question "Does the operator's trust level match the system's actual capabilities?" It does not answer the question "Is the system actively manipulating the operator's trust?" This is a significant distinction: a system could pass all CaTE trustworthiness evaluations while simultaneously producing sycophantic output that inflates operator confidence beyond warranted levels.

Revisit Triggers

  • CaTE publication of additional guidebooks or standards beyond the LAWS TEVV guidebook
  • CaTE adoption of AI safety vocabulary (sycophancy, alignment, reward hacking)
  • CaTE expansion into GenAI/LLM trust calibration (current focus is autonomous systems, not language models)
  • Publication of the DEVCOM Armaments Center trust measurement data referenced in secondary sources