c2023-4-category-failure-taxonomy

IN premise — summaries/2026/08/24/cohen-2023-ripple-effects-sR-references.md

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

Cohen et al. (2023) define a 4-category failure taxonomy for post-edit responses: correct edit, no change, abstention, and noise

Summary

This sets up a shared vocabulary for diagnosing what goes wrong when system outputs are corrected afterward: the fix landed correctly, nothing changed, the system bailed out, or the fix was pure garbage. It means any analysis of post-edit quality in this system should bucket results into one of these four outcomes rather than treating failures as a single undifferentiated blob.