R0058/2026-04-03/C001/SRC01
Roytburg & Miller (2025) — Primary quantitative study on AI safety-ethics community divide
Source
| Field |
Value |
| Title |
Mind the Gap! Pathways Towards Unifying AI Safety and Ethics Research |
| Publisher |
arXiv (preprint); accepted at IASEAI 2026 |
| Author(s) |
Dani Roytburg (Carnegie Mellon University), Beck Miller (Emory University) |
| Date |
2025-12-10 |
| URL |
https://arxiv.org/html/2512.10058 |
| Type |
Research paper (bibliometric/network analysis) |
| Origin |
Researcher-provided |
Summary
| Dimension |
Rating |
| Reliability |
Medium-High |
| Relevance |
High |
| Bias: Missing data |
Some concerns |
| Bias: Measurement |
Low risk |
| Bias: Selective reporting |
Low risk |
| Bias: Randomization |
N/A — not an RCT |
| Bias: Protocol deviation |
N/A — not an RCT |
| Bias: COI/Funding |
Low risk |
Rationale
| Dimension |
Rationale |
| Reliability |
Medium-High rather than High because: (1) this is a preprint accepted at a conference but not yet journal-published with full peer review; (2) the methodology is rigorous (null model comparisons, 500-2000 replications, p<0.001 significance tests) but the dataset is limited to 12 specific venues over 5 years. Carnegie Mellon and Emory affiliations are credible. |
| Relevance |
High — this is the primary source for the claim's specific numerical assertions. The 83.1% homophily figure and bridge concentration statistics are directly reported. |
| Bias flags |
Missing data concern: the study examines only 12 ML/NLP conference venues; researchers publishing in other venues (journals, workshops, interdisciplinary conferences) are not captured, which could systematically undercount cross-field collaboration. No conflicts of interest or funding bias identified — the authors appear to advocate for integration, which could introduce framing bias but does not affect the quantitative measurements. |
| Evidence ID |
Summary |
| SRC01-E01 |
83.1% global homophily — Figure 1 caption and supporting statistics |
| SRC01-E02 |
Top 1% of authors by degree control 58.0% of cross-disciplinary paths |
| SRC01-E03 |
Mixed papers represent 9.5% of filtered corpus; reachability analysis |