Saturday, June 30, 2018

More States Opting for Robo Grading of Student Essays.

This article from NPR on the use of software to grade student papers refers to a trend that is so appalling, I almost don't know where to begin. I was for a decade the person responsible for one of the standardized test programs in Alberta, and we were very proud of our involvement of classroom teachers at every stage of the development and marking process so that the exams would reflect actual classroom practice and the results would be normed accurately across Alberta. We trained teachers to use rubrics to mark accurately and consistently so that it wouldn't matter which marker the student got or what time of day the marking happened, or other chance factors. We did not train them to be robots, but designed rubrics that allowed them to award excellence, even when the student's answer did not easily fit within the rubric. If more than a few answers didn't fit the rubric, we changed the rubric. Routinely.

In social studies, we eliminated bias by making it explicit: when a marker found himself grinding his teeth over a student's stated opinions, instead of trying to grade it himself, he would hold the paper up above his head and shout "right-wing nut-job" or "left-wing snowflake" (as the case applied) and a suitably left or right oriented marker would happily swap it for its opposite, so that every paper was marked by a sympathetic marker. When a paper managed to offend everyone, it was taken off the marking floor and sent to a special committee of veteran markers who would grade it as a team. We went out of our way to not be robots.

This is NOT the case with many standardized tests, particularly in the US where it is private publishers rather than the ministry of education designing the exams. There are so many things wrong with most standardized testing programs, I will limit myself in this post to the observation that the purpose of most standardized testing programs is not to educate, nor to reward diligent learning, but to reproduce the class structure. Most tests reward the culture of the white upper-middle-class, and screw everybody else. They are there to explain inequality by saying, "Well you had your chance, but you only got a 58% on your test, so you deserve to spend the rest of your life in the underclass, unlike Frank here, who got 98%"(because we asked Frank questions in a way that makes the most sense to most white males, about stuff that matters to the professional/management class that Frank grew up in and had picked up on from his professional parents by the time he was 8, and because Frank had money for tutors, books, computers, if he happened to turn out a bit 'slow').

So...replacing teacher-markers with robots is perfect. Relying on software APPEARS to increase objectivity by removing the last vestiges of human intervention (what will be identified as 'bias' as these programs are being implemented) and keeps conscientious teachers from giving an "A" to a paper for its ideas when the student has written "ain't never" which is grammatically incorrect--i.e., has written her answer in the perfectly clear dialect of a to-be-suppressed population. A good teacher knows quality writing when they see it; a software program has the algorithms to suppress the underclass.

Not why I became an educator.