AI can make us more capable—or less original. After a close call on a panel discussion, I developed a simple framework for deciding when AI should extend human thinking and when it risks replacing it.
GUEST COLUMN | by Michelle Odemwingie
I was halfway through a virtual panel when it happened.
Another panelist started answering a question we’d both been prepped on, and as they spoke, I felt something cold settle in. The phrases were almost identical to the prep document I’d gotten from ChatGPT that morning. The turns of phrase, the sequencing of ideas, the framing. I was sitting there looking at my notes, thinking: you used ChatGPT for this too.
‘The phrases were almost identical to the prep document I’d gotten from ChatGPT that morning. The turns of phrase, the sequencing of ideas, the framing.’
I wasn’t caught. The panel was virtual, I pivoted, and I’ve been a comfortable extemporaneous speaker my whole life. But the near-miss wasn’t the point. The point was what I almost did: present as my own perspective something that was, in fact, the averaged output of a language model that another panelist had also queried. We had, without knowing it, nearly delivered the same talk.
That’s when I understood what was actually at stake.
I’ve used a framework for a while now that I think of as “Calculator vs. Crane.” I first encountered the crane as a frame for thinking about technology in a conversation with Dr. Sarah Johnson, who used the contrast between a crane and a loom to describe how machines encode knowledge differently. That distinction stuck with me, but I found myself reaching for a different comparison; one organized not around what machines do, but around what happens to the person using them.
Calculator or Crane?
The core distinction is simple. A calculator does something you already know how to do, faster and with more accuracy. You understand the underlying math; the calculator handles the execution. A crane does something you physically cannot do alone. One person with the right technology can lift what no human body could.
That’s not dependency. That’s extension.
MICHELLE ODEMWINGIE
The problem is that most of us, myself included for a stretch, have been using AI in calculator mode while telling ourselves it’s a crane.
I prompt AI between 50 and 100 times a day. I was a Gmail beta user in middle school, downloaded ChatGPT the week it required an invitation, and made my entire senior leadership team sign up before most of them wanted to. So when I tell you I started writing 200 words by hand every night before bed, you should understand what that admission cost me.
The panel just made it undeniable. When we outsource the things that make us distinctly human, we gain efficiency and lose the rough edges that make a perspective worth hearing. Every distinct voice, rubbed smooth by the same model, starts to sound like every other voice rubbed smooth by the same model. That is not a productivity gain. It is a loss we cannot afford.
Extension vs. Substitution
The crane version of AI is real, and I believe in it. When I took research my organization had accumulated across dozens of state landscape scans and used AI to synthesize it into policy advisory work we had no capacity to produce manually, that was crane mode. The technology extended what a small team could do. My judgment closed it. The reach was genuinely new.
But I also sat in an interview and watched a candidate use AI to generate an idea, flesh it out, and build a project plan without producing a single original thought. They were conducting an orchestra they couldn’t play an instrument in. Both the strong performers and the weak ones on my team use AI. The difference is whether they have built enough of their own instrument to hear when the output is wrong, or incomplete, or indistinguishable from what the person on the next panel is about to say.
The Question Education Is Avoiding
That distinction matters enormously for adults. It is existential for children.
When I look at AI policies coming out of elite private schools, I see institutions wrestling seriously with a specific question: how do we ensure students achieve the right cognitive development before we allow AI to augment it? The parallel to calculators is deliberate. We teach children to multiply before we hand them a calculator in fifth or sixth grade. These schools are building equivalent gateways, defined moments in a student’s learning and development when direct AI interaction becomes appropriate, rather than leaving that question open-ended.
‘We teach children to multiply before we hand them a calculator in fifth or sixth grade.’
What I see coming out of much of the public school sector is different. There are AI policies, but they are organized primarily around cheating, grading, and scoring. Those are real concerns. They are not the same concern. A policy that asks whether a student used AI to write an essay is asking a compliance question. A policy that asks whether a student has developed enough of their own reasoning to know when the AI is wrong is asking a developmental question. We need far more of the latter.
As adults, we can feel the calculator versus crane dynamic in our own work. Our brains are fully formed. We have a baseline to return to. The young people sitting in front of these screens are still building theirs.
The education sector must begin asking not just what AI policies protect students from, but what cognitive foundation we are protecting in them. Are we teaching young people to use their own cranes? Or have we found a glorified way to hand every child a calculator before they’ve learned the math?
—
Michelle Odemwingie is CEO of Achievement Network (ANet), where she advances instructional and assessment systems for under-resourced districts. A former classroom teacher, she advises education leaders on building coherent learning systems shaped by curriculum, assessment, and emerging technologies. Connect with Michelle via LinkedIn.
0 Comments