AI-Enabled Social Cyber Maneuver Detection and Creation

As social media platforms have become central to information dissemination, influence operations, and narrative shaping, understanding their role within the broader information environment is increasingly vital. The BEND framework offers a structure for analyzing online influence by identifying social-cyber maneuvers. The BEND framework was previously operationalized at the message and individual levels. In this thesis, I operationalize the BEND framework at the population and effects levels, integrate both sets of work, and align them with U.S. military doctrine and training. In doing so, I identify the critical need for complex, realistic, and scalable social media training environments. To meet this need, I introduce the AI-Enabled Scenario Orchestration and Planning (AESOP) tool, which enables planners to create training scenarios that specify events, actors, social media platform accounts, and narratives. AESOP generates synthetic templates associated with the scenario and accompanying news articles, media content, and URLs. I then present SynTel and SynX, agent-based simulation and generation tools. These tools consume AESOP-generated synthetic templates and, with support from external large language models, produce realistic and interactive synthetic social media data for X/Twitter and Telegram. These simulations replicate influence ecosystems at scale. Finally, I propose and validate a novel effects-based approach to detecting BEND maneuvers within topic-oriented groups. This technique is applied to real-world datasets to link maneuver effects to broader campaign impacts. Together, these contributions enhance our capacity to detect, evaluate, and train against influence operations — making BEND a practical analysis framework for the information environment