review-evaluates-direct-antecedents-only
IN premise
`review-beliefs` presents each derived belief with only its direct antecedents to the LLM reviewer, not the transitive chain — producing occasional sufficiency false positives when support exists one level deeper, which must be triaged manually.
Summary
When the system asks an LLM to evaluate whether a derived belief is well-supported, it only shows the immediate parent beliefs, not the full chain of reasoning behind those parents. This means the reviewer sometimes judges a belief as insufficiently supported when the real justification exists further back in the chain, requiring a human to sort out these false alarms.
Details
| Source | entries/2026/05/05/review-beliefs-in-practice.md |