AI Corrigibility · Page checks
| Severity | High | Medium | Low |
|---|---|---|---|
| Found | 1 | 2 | 2 |
| Fixed | 0 | 0 | 0 |
Overall: Article is well-structured and covers the topic comprehensively for a new encyclopedia entry.
Strengths
- Clear lead definition
- Good coverage of the corrigibility spectrum, theoretical challenges, and major approaches
- Useful "See also" section with wiki links
- Appropriate tags
Issues found
🔴 High: The paper attribution in "Background and motivation" states "2016" but the foundational Soares et al. paper was presented at an AAAI 2015 workshop. The references section correctly cites 2015; the body text should be corrected to match.
🟡 Medium: The article references the EU AI Act without noting that its enforcement is phased (2024–2027), which is relevant context.
🟡 Medium: "Corrigibility through constitutional AI and RLHF" could note that these are empirical approximations with no formal corrigibility guarantee.
🟢 Low: References section uses informal citation style rather than Agpedia’s citation system.
🟢 Low: Armstrong’s "utility indifference" (2010) attribution should be qualified as an informal working paper.