Incident File for AI Wording Denial
A summary, sequence, and factor analysis of the claim that a polished sentence was lightly edited.
INCIDENT FILE: AI WORDING DENIAL Reference: January 2023 | Status: Recurring
Summary
Someone says a polished sentence was only lightly edited after an AI tool wrote most of it.
The statement is partially accurate. Editing occurred. The characterization of the editing as "light" requires examination.
Sequence of Events
- AI tool prompted with task context.
- AI tool returns 240-word draft.
- Subject reads draft. Draft is serviceable.
- Subject edits draft. Duration: approximately eighteen minutes.
- Eleven sentences are revised. Three structural choices are changed. Two sections are reordered. The opening sentence is replaced entirely.
- Final document: 210 words. Shares subject-matter and some phrasing with original draft.
- Subject describes process to colleague as "just cleaning up the AI's wording a bit."
Contributing Factors
The help is useful enough to use and awkward enough to disguise.
AI Wording Denial adds one private manual check before any assisted sentence can be claimed. That check becomes part of the normal workflow. The work is real. The editing is real. The scope of assistance is understated.
This is not a moral finding. The document is good. The eighteen minutes of editing are a genuine contribution.
The finding is narrower: the word "lightly" does not describe an eighteen-minute revision of eleven sentences.
Current Status
Recurring. The incident log contains fourteen prior instances across multiple subjects.
Assisted work becomes routine while disclosure stays exceptional. The exception is not decreasing. This is expected.