Mastering the Media Maze: Large-Language Model Analysis of Media Communications During a Conflict Deployment

Summary: The objective of the study was to aggregate all media communications from a three-month conflict deployment of Canadian Medical Assistance Teams (CMAT), a small non-governmental EMT Type 1, to Poland and Ukraine, and to examine them for adherence to communications best practices. A media communications literature search using ProQuest and an organizational social media search were completed. Communications were categorized by date and as one of five media types: print, audio, video, authored by a volunteer, or third-party mention. Public information officer best practices were reviewed in the gray literature, and adherence to these practices was qualitatively assessed using Notebook LM, a publicly available large-language model. Between March and August 2022, there were 45 media appearances, including 17 with video footage, 10 articles authored by volunteers, 9 audio clips, 7 printed news articles, and two third-party mentions. Media correspondence was distributed among 21 distinct volunteers, with two formally media-trained volunteers responsible for 20 instances (44.4%) of the correspondence and the remainder trained by briefing alone. Overall, no overtly negative or critical events occurred, for example, breaches of ethics or security. Volunteers generally followed media orientation directions regarding communications and provided accurate information to the media. A dedicated communications specialist is needed to ensure a wide range of media, including traditional methods and social media, is created and managed, and that volunteers receive media training. Lastly, there is a constant security threat associated with divulging operational details to the press, which was addressed and successfully mitigated during our deployment.