One Small Step for Space Acquisition Doctrine
Data Analysis The study comprised 45 participants and 580 narrative-style entries. The initial strategy involved a manual review of the data (i.e., a question- by-question review of all responses from all participants). However, as the number of participants grew, it became apparent that this method was neither efficient nor effective in leveraging modern technology. Therefore, the study employed four methods to review and analyze the qualitative data: manual review, NVivo, and two Artificial Intelligence (AI) tools—OpenAI's ChatGPT 4.0 and Anthropic’s Claude 3 Opus, both of which are large language models (LLMs). Both LLMs required a premium subscription. First, the overall data set was exported from Qualtrics into an Excel file and several Word documents. Excel provided a means of consolidating all data for manual review and data traceability (based on ChatGPT or Claude results). A Word document was created for each question, containing (a) the question and (b) each participant’s responses. This method was used to ensure repeatability—in case the LLM failed or gave erroneous results—and so that each model could accurately distinguish each participant’s response.
Next, the data were coded using each LLM and NVivo. In qualitative research, the process of analyzing data is called coding. Coding is a way to annotate and categorize data. If raw data are analyzed, coding can be used to identify categories, patterns, or themes (inductive approach).
150
Defense ARJ, Summer 2025, Vol. 32 No. 2: 132—193
Made with FlippingBook - Online Brochure Maker