Literature review
Trustworthy AI in National Defense
High-consequence AI exposes the limits of treating trustworthiness as a technical property alone; it must also be governed, interpreted, and sustained by human and organizational systems.
The question
What does the research literature reveal about building, evaluating, governing, and trusting AI where uncertainty and consequence are unusually high?
What the work examines
- Governance, accountability, and the legal and ethical dimensions of AI use.
- Technical assurance, robustness, validation, and evidence of system performance.
- Human trust, calibrated reliance, teaming, and meaningful oversight.
- Ethics-by-design and the organizational conditions that support it.
Current direction
This coauthored Columbia review synthesizes 138 sources across disciplines. I use that evidence base as a high-constraint reference point for trustworthy AI in commercial, regulated, and other consequential settings—not as a boundary around the work.