Hybrid/AI Scoring: Mechanics and Implications for Instruction

 

TEA kind of announced that AI is scoring stuff for ECR and SCR when they offered the testing administrator training this month. Ok. So we found out about it in a backhanded way. Moving on. 

My little rescue dog, Joy, is trained well. For one trick only: Be cute. Then she sits. Like Joy, the AI scoring machine will do exactly what it is trained to do, regardless of what we call it. For ECR/SCR on STAAR, the scoring will do exactly what it says it will do. 

We are going to be ok. It's time. We already have the technology to hear, transcribe, and interpret English vocabulary and syntax. Phones, text messages, emails, docs, etc. It's working pretty well. (I remember when Dragon Dictation started. Didn't work too well. But it learned and got better.) 

A friend of mine saw the video I posted on facebook about this: 

HiI just came across your video you posted after meeting with your favorite English teachers. I’m not savvy to all the things your referencing in the video, but I will say…if AI allows the teachers a way to grade the students essays, I’m all for it. I would love to have all those HOURS back with my mom, my favorite English teacher. It was just our normal at the time…but looking back, she spent so much precious personal time at home “grading papers” when I was growing up.

Wow. That kind of hurt. And, we know that marking papers is actually a waste of time. See my dissertation. And I can't tell you how many hours I wasted in carrying those bags of papers home. 

So the implications for instruction are: 

1. Make sure teachers and students have access to technology that can "score" and grade papers for mechanics. Even if it costs money. This should be provided by TEA if they are using it to score. Assessment must match instruction. AI scoring resources are now instructional materials. 

2. Stop editing (grading) papers for writers. Don't do something a computer can do. 

Now - our next treat is going to be what all of this means for scoring content. More on that next.