R0044/2026-03-29/Q004/SRC02/E01¶
CMU/SEI description of CaTE mission: focuses on human understanding of AI capabilities and trust calibration.
URL: https://www.cmu.edu/news/stories/archives/2024/april/sei-and-dod-center-to-ensure-trustworthiness-in-ai-systems
Extract¶
CaTE was launched in 2023 as a collaborative partnership between CMU's SEI and OUSD(R&E). Kimberly Sablon, principal director for trusted AI within OUSD(R&E), stated: "The human has to understand the capabilities and limitations of the AI system to use it responsibly."
CaTE addresses "the dynamics of how systems interact with each other, and especially the interactions between AI and humans." It collaborates with all military branches on "measurable trust."
15 researchers are listed as involved in CaTE work.
The center's approach: "developing standards, methods, and processes for providing evidence for assurance and developing measures to determine calibrated levels of trust."
Trust calibration is defined as achieving "an ideal state in which the human places an appropriate amount of trust in machine intelligence based on its strengths and weaknesses." Overtrust occurs "when the worker trusts the automation more than it merits based on its capabilities."
JUDGMENT: The language is consistently human-focused. "The human has to understand," "the human places an appropriate amount of trust." The system's role is to be trustworthy (a property), not to adjust its behavior (an action). CaTE does not address the question: "Should the AI system itself detect and counteract user overtrust?"
Relevance to Hypotheses¶
| Hypothesis | Relationship | Strength |
|---|---|---|
| H1 | Contradicts | CaTE's descriptions are consistently human-focused without addressing system output behavior |
| H2 | Supports | Human understanding and trust calibration are the explicit focus |
| H3 | Supports | System "trustworthiness" is addressed as a property to be evaluated, not as behavior to be constrained |