AI Attitudes: Senior Service College Student Reflections on AI Use
Abstract
Artificial intelligence in military operations gets debated in the extremes. On one side, it is considered dangerous, with arguments for tight control, while the other argument counters AI is already integrated into our weapon systems, so we must learn to adapt. As the debate rages on about AI use in the military, particularly in professional military education (PME), those arguments range from complete AI integration to banning AI use due to fear of it replacing human critical thinking. Beneath this crossfire, military officers are already using AI in both their work and their military education. The more insightful question to ask at this point is this: how are senior military leaders using and thinking about AI in their daily professional work?
That question is more consequential than simply “use it or not.” The officers attending senior-level professional military education today will be the flag officers and senior executive service civilians overseeing AI procurement, integration, and employment over the next five to ten years. It is the attitudes these leaders hold about AI, not AI itself, which determines how effectively organizations adopt it. The opinions and attitudes formed in seminar rooms and online classrooms are a national security consideration.
To understand these attitudes and opinions, I conducted an exploratory review of 379 individual written reflections submitted by students in the US Army War College (USAWC) Distance Education Program (DDE). I used Gemini AI assistance for keyword identification and thematic grouping. The reflections were required as part of an assessed academic assignment. The assignment directed students to engage with a large language model (LLM) of their choice and, through prompting and iteration, make AI create a 250-word online discussion post and submit that AI-generated post for a grade. The assignment question covered operational planning and domain integration. Students needed to submit AI’s analysis and written response, not their own human-written response. The distinction matters; the design forced students to use AI and ensure that the AI output was accurate, followed the rules of the assignment question, and was at a high enough standard for grade submission. Students then provided a 50-word reflection on their experience. Collectively, these reflections provide a candid look into how experienced officers assess and interact with AI tools, some of them for the first time.
Trends
There were positive and negative attitudes regarding AI, particularly for this assignment. Students felt that AI increased efficiency, organization, and drafting while lamenting that AI created superficial and unnuanced answers, often with inaccuracies or hallucinations. The most common platforms used were GenAI.mil (version Gemini 2.5 at that time), ChatGPT (5.1 and 5.2), Microsoft Copilot, and NotebookLM (now Gemini Notebook). The reasoning behind the selection of certain platforms largely centered on accessibility on government machines, desire for higher compute power, and desire for more trusted and grounded results. Approximately 25-30% of students used multiple AI tools for cross-verification of the output and for functional AI specialization (Grammarly for grammar, ChatGPT for generation). Although the students were required to submit AI’s full written response for this assignment, many reflected on how they typically use AI: summarization of material, drafting/outlining, brainstorming, and as a Socratic partner. There were three main themes derived from the reflections that support the identified trends.
AI for Speed
The Junior Staff Officer Paradigm
The second theme was particularly interesting. Students consistently presented the same professional metaphor, unprompted, across hundreds of individual reflections: AI works like a productive, but very junior, staff officer. This staff person, AI in this case, is useful for generating drafts of products or processes, but requires extensive direction. “Never trust the first output from your Action Officers and never give the first draft to your boss without reviewing for accuracy,” one student wrote. In this case, AI was that staff officer.
This paradigm provides significant operational implications and future PME considerations. Students actively engineered their prompts and iterated toward specified results, but no student was satisfied with the first AI output, nor was there confidence that AI provided enough analysis or nuance to the assignment. From this perspective, it is productive to intuit that using AI necessitates certain competencies in communication and direction. Senior leaders must develop skills in providing useful directions to produce a meaningful result from an AI or a subordinate. Ultimately, the leader is in the loop but needs assurance that their efforts will produce the intended output.
Commenting is available to active SWJ members and staff.