Defense Acquisition Research Journal #109

https://www.dau.edu

TABLE 3. SENTIMENT ∆ Sentiment ∆ (ChatGPT vs Claude)

Question

Positive

Neutral

Negative

Q7

6%

-9%

3%

Q8

-4%

7%

-3%

Q9

0%

-1%

1%

Q10

31%

-42%

11%

Q11

7%

3%

-10%

Q12

-3%

9%

-6%

Q13

11%

18%

-29%

Q14

11%

-23%

12%

Q15

8%

-23%

15%

Q16 Q17

-4%

-6%

10%

Q18

2%

-9%

7%

Q19

2%

-2%

-1%

Q20

-9%

-1%

10%

Space Acquisition Tenet Alignment This section briefly discusses the overall quantitative alignment analysis conducted by ChatGPT and Claude. On average, each question aligned with their respective tenet between 60% to 83%. This methodology was used as a starting point for the qualitative question-by- question sentiment and alignment analysis, which will be discussed in a later section. Tables 4, 5, and 6 summarize the alignment of questions and tenets across both models and present their averages. If the model provided a range, the lower bound was used as a conservative estimate.

163

Defense ARJ, Summer 2025, Vol. 32 No. 2: 132—193

Made with FlippingBook - Online Brochure Maker