2 ms·
Inside the essay-scoring business
- _delirium 16y agoSomething I've long wondered, but which doesn't seem to be testable given currently available public data-sets: If we're going to be doing this sort of large-scale, rubric-based grading of essays, perhaps it should actually be done by computers? Given some training data, I would hypothesize that a modern statistical classifier could assign an essay a score with at least as good accuracy as a human employee doing such huge volumes of essays. It would also be more consistent, whereas it seems like there's a huge amount of person-to-person variance in the current scoring process.