Finding Deepfakes: A Tabletop Exercise About AI, Decisionmaking, and Algorithmic Performance

Authors

Andrea Brennan, In-Q-Tel Laboratories
Gwyneth Sutherlin, College of Information and Cyberspace, National Defense University
Lisa Pagano-Wallace, Office of the Director of National Intelligence
Hermie Mendoza, U.S. Coast Guard Atlantic Area C5I and Security Division (LANT-6)

Abstract

This article explores a first-of-its-kind tabletop exercise designed to expose U.S. military and national security professionals to the real-world challenges of using AI tools like deepfake detectors in high-stakes decisionmaking. Set in a fictional scenario involving a flawed AI system, participants confronted algorithmic bias, technical uncertainty, and ethical dilemmas, revealing how AI limitations can erode trust in intelligence products. Drawing from student and faculty reflections, the article highlights the importance of preparing future leaders to scrutinize and manage AI performance failures not only as technical bugs but as complex, systemic issues with social and operational consequences. The exercise underscored the need for a more holistic, human-centered approach to AI in defense contexts.

Recommended Citation

Andrea Brennan, Gwyneth Sutherlin, Lisa Pagano-Wallace & Hermie Mendoza, "Finding Deepfakes: A Tabletop Exercise About AI, Decisionmaking, and Algorithmic Performance," Joint Force Quarterly 118 (3rd Quarter 2025), 49-55, https://digitalcommons.ndu.edu/joint-force-quarterly/vol118/iss3/8.