STUDY METHODOLOGY
Sentiment Analysis
Inputs
The source census contains 256 unique articles from Spectator’s University News Administration archive in matched January 1–August 13 windows in 2024 and 2026. The analysis includes 51 records whose headline or public description matched the federal-pressure terms below. For each record, one blinded ChatGPT pass received only the headline and public description or dek.
Federal-pressure selection
A record entered the analysis when its headline or public description matched at least one published term: Trump, Congress, House committee, federal, Title VI, Education Department, White House, ICE, funding cut. Sentiment scores were not used to determine whether a record was selected.
Why these terms?
These terms were selected to identify coverage where federal pressure on Columbia was explicit in the headline or public description rather than inferred. The list captures federal political actors and institutions—Trump, Congress, House committee, White House, and the Education Department—federal legal or enforcement references such as Title VI and ICE, and direct pressure language such as “federal” and “funding cut.” Together, they provide a transparent rule for identifying coverage in which federal pressure was directly present in the supplied text.
Question and scale
ChatGPT was asked: “Based only on the supplied text, how does the article frame Columbia administrators or trustees?” It returned one score, a confidence level, and a short text-grounded rationale.
| Score | Meaning |
|---|---|
| -2 | strongly critical of administration or foregrounds a serious accountability challenge |
| -1 | mildly critical or skeptical toward administration |
| 0 | neutral, mixed, descriptive, or insufficient evidence of directional valence |
| 1 | mildly favorable or substantially centers an administration frame |
| 2 | strongly favorable, promotional, or largely uncontested institutional framing |
Samples and result
Baseline: Federal-pressure records, January 1–August 13, 2024 (n=25)
Comparison: Federal-pressure records, January 1–August 13, 2026 (n=26)
The raw score was divided by two to create a −1 to +1 scale. The mean changed from -0.380 to -0.038, a difference of +0.342.
Statistical result: The two-sided Welch test gives p=0.0366 with a 95% confidence interval from 0.022 to 0.661. At the conventional 0.05 threshold, the 2024–2026 federal-pressure comparison is statistically significant.
Reproducibility
The Excel workbook contains all analysis inputs, term matches, scores, confidence ratings, rationales, test results, and original article links. The PDF summarizes the method and result.