Fascinating results! I'd like to know more about the choice to do human scoring with the pre-test and digital scoring with the post test - was there a specific reason (maybe pragmatic/time concerns?) why you didn't do human and digital scoring at both time points? Thanks for your work and your writing. I always learn a LOT from your work.
It was mainly a pragmatic decision to manage teacher workload. We will follow up with some human judging to check the agreement. One advantage of using AI is there is no possibility of bias whereas a human may recognise that a piece had been redrafted.
Fascinating results! I'd like to know more about the choice to do human scoring with the pre-test and digital scoring with the post test - was there a specific reason (maybe pragmatic/time concerns?) why you didn't do human and digital scoring at both time points? Thanks for your work and your writing. I always learn a LOT from your work.
It was mainly a pragmatic decision to manage teacher workload. We will follow up with some human judging to check the agreement. One advantage of using AI is there is no possibility of bias whereas a human may recognise that a piece had been redrafted.
Thanks for the speedy reply!
d=.29 in 2 lessons? That's impressive because that's the normal gain in 1 year.