The Missing Stake
what we lose when AI does the work - and what the in-class assignment gets wrong
Last week’s blog post was written by AI. I came up with the idea, added several examples, cleaned up the paragraphs and transitions, ran it through my own “make sure this doesn’t sound like AI” prompt, and tied the post back to themes I have been writing about all year. But it was not mine; I didn’t really have skin in the game. My job was to suggest, iterate, review, and clean up around the edges. And while the post was accurate and readable, I didn’t feel the burden of responsibility - as I do now.
When we outsource key parts of our work to AI, we lose something. Not just our unique writing voice - which AI can easily approximate - but our thought process, our perspective, and our expertise. If that’s true for you or me, it’s at least as true for a student who has been handed an assignment they know a bot can complete in thirty seconds. The question isn’t whether they’ll use it; it’s whether the assignment gave them any reason not to.
Schools have noticed this, and the most common response has been to bring back the in-class, handwritten assignment - put students in a room alone, under pressure, without tools, and remove any ambiguity about who did the work. Teachers aren’t wrong that it works. The problem is whether or not that’s the best solution. If AI allows students to produce work that meets or exceeds grade-level expectations, some teachers will conclude that the supervised test is the only remaining guarantee of authentic effort. That would be a loss - because technology should be expanding what students can do, not pushing schools toward a narrower definition of what counts.
Preparing young people for a world where AI handles routine cognitive work by pretending AI doesn’t exist is a strategy that is already failing. As I’ve found out myself, the goal isn’t to increase or decrease the friction, but to make sure it’s the right kind.
What actively undermines education is handing off the parts of the work where learning actually happens. When students connect examples to concepts, find the right words for a complicated idea, discover that their first argument has a hole in it, or have to make their case to a skeptical classmate - that’s the goal, not just a means to an end. The messy, roundabout, often frustrating work of understanding is exactly what school is for, and it’s the first thing to disappear when AI takes the wheel.
Every assignment has a variety of tasks - some are administrative, some are mechanical, and some require genuinely personal input from the student. AI belongs in the first two categories. The useful question isn’t whether to allow AI, but where in the assignment the student’s input matters. Where are you asking them to do something that requires their own thinking - and where have you accidentally left an easy out? What’s the one contribution or insight in this assignment that can’t be handed off without diminishing the point of the whole thing?
That question will produce something more useful than a policy: a clearer sense of what you’re actually trying to teach, and whether your current practice protects it. It doesn't require a total curriculum rewrite or a return to the days of paper and pencil; just a closer look at where the human element exists in the assignments you already have.
A few years ago we hiked the West Rim Trail at Zion National Park - up Walter’s Wiggles to Scout Lookout, and the final push to Angels Landing. It was a warm spring day, but there was still snow and ice on the trail, so we rented spikes for our boots. The last half mile goes up a steep sandstone ridge with chains bolted into the rock. You need spikes and chains because the exposure is stark, the footing requires attention, and there’s no version of that climb that doesn’t ask something of you - whether you’re 30 or 60. But the view from the top, and the simple lunch we ate there, were worth every step. Not despite the difficulty, but because of it.
That’s what the best assignments do. They’re designed so that the critical parts can’t be faked; so the human involvement is real. AI already has a place in that design - clearing the clutter, handling the scaffolding, and reducing administrative burden. The challenge is to use it without removing the climb - or the reward.
See you in September.
Cheers,
-Rick