These visualizations are experimental and still being refined. Designs and data presentation may change.
Governance-Severity Relationship
Testing whether more governance correlates with fewer or less severe incidents. Toggle the Y-axis between incident count and expert BAU severity to see different perspectives on whether governance is associated with harm reduction.
How to read this
Each dot is one AI risk subdomain. The x-axis shows governance coverage (weighted score or raw document count), and the y-axis shows either incident count or expert BAU severity. Toggle between metrics to explore different hypotheses. The trend line shows the overall linear relationship — a negative slope would suggest governance reduces harm.
Key Takeaways
- 1.Governance score vs. incident count: r = 0.36 (weak). More-governed subdomains tend to have more incidents (governance follows harm).
- 2.Governance score vs. expert BAU severity: r = 0.21 (weak). Higher-severity subdomains tend to attract more governance.
- 3.4.3 Fraud, scams, and targeted manipulation has the largest incident-to-governance ratio (353 incidents, score 193.5).
- 4.6.5 Governance failure has the highest governance-to-incident ratio (score 341.5, 2 incidents).
Governance score weights enacted and proposed documents. Expert severity is BAU weighted mean (1–5).