Stop testing students. Let them do the job.

Years ago, when I was the tech coordinator for the Winneconne school district, a student walked into my class who had already failed a pile of others. The read on him was simple, and everyone shared it: lazy, checked out, didn't want to do the work.

He became one of the best students I had.

The class was Technology Innovation Problem Solving, and it ran differently than the ones he'd been failing. No worksheet, no chapter to memorize, no test at the end. Students picked a real problem in the real world and had to figure out how to solve it. I wasn't there with the answers. I didn't have them. I was there to guide.

He picked the district's scheduling system. He wanted a better way to schedule students, and he pulled together a team to build it. They put together a full plan — timetables, a budget, what it would cost, how you'd actually pay for it — and presented it to the school board. The board didn't adopt it. But the plan was rich and well thought out, the kind of work most people in that room never believed this kid could produce. It floored the teachers.

Think about the last time something actually clicked for you. I'd bet it wasn't the day you passed a multiple-choice test. It was after you tried something, failed, tried it a different way, and kept going until you could do it for real. You can gain knowledge from reading. You can recall facts on a quiz. But learning you can apply, especially at work, comes from doing the thing and getting it wrong first.

I started my teaching career the other way around. Multiple-choice quizzes and rote memorization, that was the whole toolkit. And I kept re-teaching the same material, because students would cram, pass, and let it evaporate. That's what an objective test trains you to do. You cram for it, you pass, and you move on. Most of that knowledge never got applied, so it never really stuck.

That student in Winneconne is why I've built courses the same way ever since, and it's what we ship at WGU today. Our students still learn the material, but they're assessed on artifacts — the actual outputs they'd produce on the job. Things they can put in a portfolio. Things they can talk about in an interview. We make the scenario as close to the real thing as we can, so a student has to apply everything they've learned and solve problems along the way.

Part of that scenario now is AI. Students work with AI personas built to behave like real coworkers and stakeholders, including the difficult ones. They have to communicate with a frustrated person and bring the situation back around. They have to pull requirements out of someone who gives vague answers, ask good follow-up questions, and work through messy information until they can figure out what's actually being asked. That's the job. That's what we're assessing.

And when a student gets it wrong, they revise and resubmit. This is the part people push back on. Letting students retake until they pass, the objection goes, is just grade inflation.

Here's how I see it. Learning is competency-based. If you can prove you can do the job — apply the skill against a real scenario and deliver the work — then you've proven what you can do. It doesn't much matter whether it took one attempt or three. You don't do things right the first time unless you've done them before. Resubmitting isn't inflation. It's how learning works.

And early on, it's working. The work got harder, not easier. Students are being asked to do real tasks now, not bubble in answers. Even so, we're seeing early signs that drop rates are improving and overall pass rates are climbing, and students are telling us they enjoy the experience more. A lot of that is the resubmit model itself. A student who has to revise and turn it in again often ends up doing better, and gets back on track faster, because now they've actually practiced the skill instead of failing once and walking away.

We're still figuring out how to make every scenario as real as the one that lit up that kid in Winneconne. Some courses are closer than others, and some of the AI personas still need work. But I'm not going back to testing students on what they can memorize. I'd rather find out what they can actually do — the way that student's teachers finally did.