Skip to content

R0058/2026-04-03/C001/H1

Research R0058 — Candidate evidence test
Run 2026-04-03
Claim C001
Hypothesis H1

Statement

The claim is accurate as stated: AI research communities show approximately 83% homophily between safety and ethics subfields, and only approximately 1% of authors bridge the divide between these communities.

Status

Current: Eliminated

Supporting Evidence

Evidence Summary
SRC01-E01 Roytburg & Miller report 83.1% global homophily, confirming the first part of the claim

Contradicting Evidence

Evidence Summary
SRC01-E02 The paper does not state that "only 1% of authors bridge the divide." It states that the top 1% of authors by degree control 58% of cross-disciplinary paths — a fundamentally different claim about network centrality, not bridging authorship rate

Reasoning

While the 83% homophily figure is confirmed, the "1% bridging" component is a mischaracterization of the source finding. The paper's actual claim is about the concentration of bridging influence among high-degree nodes, not about the total percentage of authors who bridge. Additionally, the paper reports that mixed papers represent 9.5% of the filtered corpus, suggesting bridging activity is more common than "1%" implies. Because the claim as a whole requires both components to be accurate, H1 is eliminated.

Relationship to Other Hypotheses

H1 fails because of the inaccuracy in the second component. This supports H2 (partially correct), since the first component is confirmed while the second is a mischaracterization.