LO
CapabilityCognitive automation2024-09-22
Experts largely preferred text edited by other experts over automatic edits, in a study of 1,057 AI-written paragraphs edited by professional writers
Editor and proofreaderoccupation page →
Event date / reported
2024-09-22
Evidence stage
CapabilityA demo, benchmark or paper shows the task can be done. Updates what the technology can do — not what employers will do.
Tasks this bears on
Copyediting for accuracy and style
Editing text line by line for clarity, style and accuracy.
Being augmented✓ Evidence-backed
Structural and developmental editing
Shaping a manuscript's structure, argument and voice with its author.
Still human-led≈ Platform inference
Where this applies
A study in which professional writers edited 1,057 paragraphs written by language models according to a taxonomy of problems such as clichés and unnecessary exposition. The authors report that a large-scale preference annotation confirms experts largely prefer text edited by other experts, while automatic editing methods show promise. It concerns literary and creative text.
What this means
Editing AI text well is itself skilled work that experts still do better than machines.
What it does not yet show
One study of literary paragraphs; it does not measure editing jobs.
What you can check
Open the arXiv paper 2409.14509 and find "experts largely prefer text edited by other experts".
Does it change the assessment?
No — and this stage does not move it either. A "Capability" record is real evidence, but it does not upgrade a task judgement on its own. The 2 linked judgements above stand where they were.
Source
Chakrabarty, Laban and Wu — Can AI writing be salvaged? Mitigating Idiosyncrasies and Improving Human-AI Alignment in the Writing Process through Edits, arXiv:2409.14509 (ACM CHI 2025; submitted 22 Sep 2024) · verified 2026-09-30 · Claude (VOLO agent) · interpreted 2026-09-30 · Claude (VOLO agent)
Primary source — published by the party that did this, or the authority of record. No co-signature needed.