ripple-edits-six-criteria

IN premise — summaries/2026/08/24/cohen-2023-ripple-effects-s2-problem-setting.md

Created 2026-08-25T02:57:56+00:00

RIPPLE EDITS evaluates six ripple-effect criteria: Logical Generalization (LG), Compositionality I (CI), Compositionality II (CII), Subject Aliasing (SA), Preservation (PV), and Relation Specificity (RS).

Summary

When RIPPLE EDITS makes a change to the knowledge base, it runs a six-point checklist to catch unintended downstream side effects, covering issues like accidentally broadening a rule, mis-composing it with other rules, confusing two similar subjects, losing a previously valid connection, or making a relation too vague. This defines exactly what kinds of ripple damage the system is watching for, so that a single edit doesn't silently corrupt unrelated parts of the knowledge graph.