Abstract:

The rapid proliferation of synthetic and manipulated media poses a growing threat to mission resilience, information operations, and public trust. Adversaries exploit advanced technologies to disseminate falsified media, undermining decision-making processes and operational effectiveness. This paper explores emerging methodologies and innovative concepts for enhancing human-AI collaboration in detecting and countering synthetic media, with a focus on applications in defense, homeland security, and broader operational contexts.

Building on insights from a comprehensive evaluation of synthetic media detection systems, we highlight novel strategies for integrating human expertise with AI capabilities to advance training and operational effectiveness. The evaluation spanned over 111 tasks involving multi-modal data to assess manipulation detection and localization, generator attribution, and intent characterization. These tasks simulated real-world threats through curated datasets, enabling the identification of system strengths and areas requiring further development.

Findings emphasize the complementary roles of human and AI contributors: humans excel in contextual reasoning and nuanced judgment, while AI systems provide unmatched speed and scalability. Case studies piloting these approaches in operationally relevant environments revealed promising pathways for training the workforce of the future, improving decision-making under pressure, and fostering trust in human-machine teaming. We further explore challenges in workflow integration, trust calibration, and human-centric design, offering actionable recommendations for advancing training scenarios and simulation environments.

This work underscores the transformative potential of human-AI collaboration in tackling emerging threats and demonstrates how modeling and simulation can accelerate the development of groundbreaking capabilities to secure the information environment.

DISCLAIMER: The research described herein is sponsored by the Defense Advanced Research Projects Agency (DARPA, Contract No. 47QFLA22F0137). The views, opinions and/or findings expressed are those of the authors and should not be interpreted as representing the official views or policies of the Department of Defense or the US Government.

Distribution Statement “A” (Approved for Public Release, Distribution Unlimited)

Keywords: EVALUATION;SIMULATIONS;SYNTHETIC;TRAINING

Menu

Peter Bautista