Skip to content

R0058/2026-04-03/C001

Claim: AI research communities show 83% homophily between safety and ethics subfields, with only 1% of authors bridging the divide.

BLUF: The 83% homophily figure is confirmed (83.1% measured by Roytburg & Miller, 2025). However, "only 1% of authors bridging the divide" mischaracterizes the source: the paper reports the top 1% of authors by network degree control 58% of cross-disciplinary paths, a finding about concentration of bridging influence, not the total number of bridging authors.

Probability: Likely (55-80%) | Confidence: High

Correction needed: Replace "only 1% of authors bridging the divide" with a more accurate characterization such as "the top 1% of authors by network centrality control 58% of all cross-disciplinary connections" or "cross-field connectivity is concentrated among a small elite of bridge authors."


Summary

Entity Description
Claim Definition Claim text, scope, status
Assessment Full analytical product with reasoning chain
ACH Matrix Evidence x hypotheses diagnosticity analysis
Self-Audit ROBIS-adapted 5-domain audit (process + source verification)

Hypotheses

ID Hypothesis Status
H1 Claim is accurate as stated Eliminated
H2 Partially correct — 83% confirmed, 1% mischaracterized Supported
H3 Claim is materially wrong Eliminated

Candidate Evidence

ID Source URL Status
CE01 Roytburg & Miller (2025) https://arxiv.org/html/2512.10058 Scored as SRC01

Searches

ID Target Results Selected
S01 Candidate evidence fetch 1 1
S02 Direct search for primary paper and related studies 20 3
S03 Contradicting evidence search 10 1
S04 Independent bibliometric studies 10 2
S05 Bridge workshops and venues 20 2

Sources

Source Description Reliability Relevance
SRC01 Roytburg & Miller (2025) — primary quantitative study Medium-High High
SRC02 Qiu, Cheng & Huang (2025) — independent bibliometrics Medium Medium
SRC03 Mehrotra et al. (2025) — AIES/FAccT scoping review Medium Medium
SRC04 NeurIPS 2019 "Minding the Gap" workshop Medium Medium

Revisit Triggers

  • Roytburg & Miller journal publication: If the paper is published in a peer-reviewed journal with revised figures (currently a preprint accepted at IASEAI 2026), the specific numerical claims may change.
  • Replication study: If an independent team conducts a similar network analysis of AI safety-ethics co-authorship using different venues or timeframes, the 83.1% figure may be confirmed, revised, or contradicted.
  • IASEAI 2026 proceedings: When the conference proceedings are published, check for any reviewer-requested revisions to the methodology or figures.
  • Significant institutional changes: If major AI research organizations (e.g., DeepMind, OpenAI, Anthropic) establish formal safety-ethics integration programs, the homophily measurement could decrease significantly.