Skip to content

R0058/2026-04-03/C001/H2

Research R0058 — Candidate evidence test
Run 2026-04-03
Claim C001
Hypothesis H2

Statement

The claim is partially correct: the 83% homophily figure is accurate, but the "1% of authors bridging the divide" mischaracterizes the source finding. The actual finding is that the top 1% of authors by network degree control a disproportionate share of cross-disciplinary connections, which is a statement about concentration of bridging power, not about the total number of bridging authors.

Status

Current: Supported

Supporting Evidence

Evidence Summary
SRC01-E01 Confirms 83.1% global homophily — over four of five collaborations are within-group
SRC01-E02 The top 1% of authors by degree control 58.0% of cross-disciplinary paths; the top 5% control 88.1%. This is about concentration, not about the absolute number of bridging authors
SRC01-E03 Mixed papers represent 9.5% of the filtered corpus, indicating more cross-field activity than "1%" implies
SRC02-E01 Independent bibliometric study finds 94% of AI ethics institutions are in the largest connected component, suggesting the broader field is connected even if safety-ethics specifically is siloed
SRC04-E01 The existence of NeurIPS workshops dedicated to bridging fairness/ethics/safety confirms the recognized gap but also shows active efforts to bridge it

Contradicting Evidence

Evidence Summary
None No evidence was found contradicting the partial-correctness assessment

Reasoning

The evidence strongly supports this hypothesis. The 83.1% figure from Roytburg & Miller directly confirms the first part of the claim. However, the paper's finding about the top 1% of authors is about the concentration of cross-disciplinary influence among a small elite, not about the total percentage of authors who cross the divide. The 9.5% mixed-paper rate further demonstrates that bridging activity, while uncommon, is more prevalent than a "1%" framing suggests. Independent bibliometric work and the existence of dedicated bridging workshops corroborate both the reality of the divide and the inaccuracy of characterizing it as "only 1% bridging."

Relationship to Other Hypotheses

H2 is the best-supported hypothesis. It subsumes the true portion of H1 (the 83% figure) while explaining why H1 fails (the 1% mischaracterization). It is incompatible with H3 because the core phenomenon (high homophily, siloed communities) is well-evidenced.