E03¶


Research	R0058 — Candidate evidence test
Run	2026-04-03
Claim	C001
Source	SRC01
Evidence	SRC01-E03
Type	Statistical

Mixed papers represent 9.5% of filtered corpus; reachability analysis shows structural barriers

URL: https://arxiv.org/html/2512.10058

Extract¶

Mixed papers (those spanning both safety and ethics) represent 9.5% of the filtered corpus
After 5 hops in the co-authorship network, only 16.9% of safety-ethics author pairs are connected
Expected random connectivity after 5 hops: 21.5% (p<0.001)
The 4,564 authors with 2+ papers form the analyzed network

The 9.5% mixed-paper rate indicates that roughly 1 in 10 papers touches both communities — far more than the "1%" of bridging authors suggested by the claim.

Relevance to Hypotheses¶

Hypothesis	Relationship	Strength
H1	Contradicts	The 9.5% mixed-paper rate contradicts the "only 1% bridging" characterization. Even accounting for the difference between papers and authors, the actual cross-field activity level is substantially higher than 1%.
H2	Supports	The reachability analysis confirms the divide is real (16.9% vs. 21.5% expected) while the mixed-paper rate shows the "1%" framing understates bridging activity
H3	Supports	The reachability deficit (16.9% vs. 21.5%) confirms the divide but its modest magnitude (4.6 percentage points) could be read as the divide being overstated. However, the homophily measure is more directly relevant than the reachability measure.

Context¶

The reachability analysis provides a complementary measure to homophily. While 83.1% homophily measures edge-level in-group preference, the reachability analysis measures whether researchers can reach each other through the network at all. The below-random reachability confirms the structural divide but the magnitude is modest.

Notes¶

The 9.5% mixed-paper figure is important for assessing the "1% bridging" component of the claim. Papers with mixed methods or topics spanning both fields are nearly ten times more common than the claim suggests for bridging authors.