One Small Step for Space Acquisition Doctrine
for future doctrine development research. Table 2 shows question-by- question sentiment across both models and their respective averages. Note that the color trend for positive sentiment was greener as the percentage rose. The neutral and negative sentiments used the inverse color trend; lower percentages were represented by green, and higher percentages by yellow.
TABLE 2. SURVEY SENTIMENT Sentiment Analysis
ChatGPT 4.0
Claude Opus
Average
Positive Avg
Neutral Avg
Negative Avg
Question Positive Neutral
Negative Positive Neutral
Negative
Q7
73% 20% 7% 67% 29% 4% 70% 25% 6%
Q8*
30% 50% 20% 34% 43% 23% 32% 47% 22%
Q9
62% 17% 21% 62% 18% 20% 62% 18% 21%
Q10*
62% 18% 20% 31% 60% 9% 47% 39% 15%
Q11
35% 40% 25% 28% 37% 35% 32% 39% 30%
Q12
65% 25% 10% 68% 16% 16% 67% 21% 13%
Q13*
40% 40% 20% 29% 22% 49% 35% 31% 35%
Q14
35% 20% 45% 24% 43% 33% 30% 32% 39%
Q15
70% 10% 20% 62% 33% 5% 66% 22% 13%
Q16 Q17
70% 15% 15% 74% 21% 5% 72% 18% 10%
Q18
15% 15% 70% 13% 24% 63% 14% 20% 67%
Q19 69% 22% 9% 67% 24% 10% 68% 23% 10% Q20 80% 10% 10% 89% 11% 0% 85% 11% 5% Avg Total 54% 23% 22% 50% 29% 21% 52% 26% 22% * If positive, negative, or neutral data varied by more than 20 percentage points, data were re-run. Two trends emerged from the data. Questions 8, 10, and 13 initially had a difference of more than 25 percentage points, and the models were re-run (see model delta in Table 3). After the second iteration, Question 8’s disparity was reduced, while Questions 10 and 13 remained high. Question 10 showed notable differences between positive and neutral responses between models. ChatGPT interpreted responses more positively, while Claude interpreted them more as neutral. Question 13 showed the most significant variance in negative sentiment. Lastly, Questions 14 and 15 had high variations in neutral sentiment. The sentiment analysis clarified participants’ outlooks and the effectiveness of the questions. Question sentiment is further discussed later in this section.
162
Defense ARJ, Summer 2025, Vol. 32 No. 2: 132—193
Made with FlippingBook - Online Brochure Maker