I’ve worked with David VanStory for some time now, and he’s thought more about Direct Instruction than almost anyone I know. I asked him to write up what he learned building error correction into CrushAP. Here’s David. – Zach IntroductionMost learning apps think they correct students’ mistakes. Very few actually do. The industry standard is to show the student the right answer and move them on to the next question. Everyone accepts this as the way things are, but it doesn’t have to be. We built corrections that way too, at first. I’m (David) an engineer building CrushAP, an AP Physics app based on Direct Instruction. After a lot of trial and error, a DI expert, Jon Owen, showed us that the problem was solved decades ago, in a procedure today’s learning apps have mostly ignored. Here’s how we adapted DI’s correction procedure for our app. Corrections matter more than most people thinkA tenet of Direct Instruction is that mistakes should be corrected immediately. They are so important that NIFDI made a protocol that details step by step how to correct students. The clearest version of the DI correction procedure is the “7 step correction procedure for verbal tasks”. The core of the procedure is this: the mistake is corrected immediately, then students repeat the task until it’s firm, with increasing delays. Students must prove they have corrected their mistake by giving correct responses before they are allowed to move on.
The key benefit of the DI correction procedure is that it ensures:
Here’s what a student actually does after getting a question wrong, before and after we adopted the DI correction procedure.
Students get the practice they need right away, at the moment they need it, instead of carrying a mistake forward. If they don’t have to correct it, they won’t.When a student made a mistake, we would show the full solution for them to read, and then move them on. Unfortunately, we found that students would just click “next” immediately, not reading it. It wasn’t until after many iterations of our own correction procedures that we realized that if a student doesn’t have to correct their mistake, they won’t. This is not a dig. It’s human nature. We tried several correction procedures:
All of these solutions were reasonable iterations, but they were trial-and-error designs that did not have the right balance of efficiency and proof of understanding. We didn’t know that the hard work had already been done: a tried and tested DI correction procedure already existed. Over time, we learned two things:
I sought the advice of a DI expert, Jon Owen, and he suggested I would benefit from looking at DI’s correction procedure. He was right. Adapting the procedure for an appSo, our mistake had been that our corrections did not follow DI principles. Students weren’t repeating the task with increasing delays between repeats. Here is how we have made our corrections align with DI principles. It is important to note that the 7 step DI correction procedure was designed for verbal tasks, with a teacher and a group of students. Our app is not a classroom, and that changes things:
So we didn’t try to replicate the exact procedure which was designed for verbal tasks in the classroom. We took the core of the procedure (correct the mistake immediately, repeat the task until it’s firm, test it again after a delay) and built a 4 step version that fits our app. Table 1 - How the DI Correction Procedure maps into CrushAP app
Step 1 - (DI Step 1) correct the mistake (in this case, prompt the rule)
Step 2 - (DI Step 2) repeat the questionThe student repeats the task. However, getting the answer right on the repeat isn’t enough on its own. A student can pick A because they remember it was A. So after the repeat, we ask: ‘How do you know?’. This was one of Engelmann’s key DI techniques when teaching rules. The student has to pick the reason, ‘Wider spacing means faster motion’. That tells us whether they’re applying the rule or just recalling the answer.
Step 3 - (DI Step 3) go back and then repeat the part - quick replayThe student answers an earlier question, then redoes the one they missed - we call this “quick replay”. This adds a short delay before they have to prove they get it.
Step 4 - (DI Step 7) A variant at the endAt the end of the lesson the student gets a “variant” of the question they missed: the same type of question with a different scenario. This does two things. It’s a delayed test, so they have to show they still know the rule after a longer gap. And because the scenario is new, they can’t just answer from memory - they have to apply the rule. Right now the variant comes at the end, but in longer exercises it could come sooner.
What changedThe biggest change is that a mistake now leads to more practice instead of straight to the next question. Before, a student who got a question wrong saw the answer and moved on, and we had no way of knowing whether they attended to the correction. Now every missed question is corrected and asked again right away, and about 9 in 10 of our students get it right on that retry. More than 8 in 10 still get the question right when it comes back at the end of the lesson (Student Data, August - October 2026). It also changed what we can see. Because the student has to answer again, every correction now tells us whether they actually learned from the mistake. ConclusionShowing a student the right answer isn’t the same thing as correcting a mistake. If a student doesn’t have to correct their mistake, they won’t. The DI correction procedure fixes that by making the correction part of the work: the student repeats the task, proves they get it, and proves it again later. It took us a lot of iterations to adapt the classroom version to work in an app, and there’s more to do. Next, we’re adapting the correction to the type of error, whether it’s a discrimination, strategy, or production error. But the principle is the same, and our students are getting the practice they need at the moment they need it. Thanks to Marcy Stein for reviewing drafts of this piece. Zach Groshell is free today. But if you enjoyed this post, you can tell Zach Groshell that their writing is valuable by pledging a future subscription. You won't be charged unless they enable payments.
|
No comments:
Post a Comment