Three thoughts on students using LLMs to do their thinking.
Of course, I’m not in favor of it. But it’s been hard to articulate why in a way that students would understand it.
Think I found a good analogy for this in chess.
If you watch chess videos, great moves often appear self evident after the player has made them, right?
Like, you don’t think much about it—“anyone can do that”.
But, when you play and YOU have to make a move in that same position, all of a sudden it’s not so clear anymore.
It’s the same effect with LLMs—after the LLM has “done the thinking”, it’s easy to agree with it! Like with the chess example, you’d be saying, “oh yeah, I could have written this”, or “yeah, that’s what I thought”.
But that’s just an illusion. It’s “after the fact” illusion. And students (or anyone else) can easily fall victim to it.
The second reason I don’t like students using LLMs to do their thinking is that it limits the scope of their ideas. When given a “blank page”, students can fill it with any kind of solution. However, when they use an LLM, the LLM dictates the solution space that the student even considers. This is a form of anchoring bias—the LLM anchors (and therefore limits) the student’s thinking. Instead of thinking about novel solutions, the student thinks, “I agree” or “I disagree” with the LLM’s solution. This is the wrong starting point.
Third reason is the lack of basic skills. A lot of teachers say, “it’s okay to let students use LLMs if the student checks the quality of the LLM’s output”.
I disagree with this, for the reason that students are *incapable of distinguishing quality*. That is, before they know how to produce quality themselves, they cannot say if an LLM-produced piece is of high quality. However, they will try and say, “yeah, it’s well written, it looks good.” And then they return it.
This explains why so many teachers get a surrealistic experience; they are flabbergasted, thinking “oh, I asked the students to check but they obviously didn’t!”. This is wrong thinking: the student DID check but they just didn’t know how to do it properly!
Imagine you, a teacher not knowing anything about car mechanics, to be asked to check if a repair robot has done a good job on fixing an engine. You would have no clue whatsoever.
It’s similar with students. Until they learn the basics of *logical thinking* and *expressing logical thinking in writing*, plus of course domain expertise on what they are writing about, they cannot evaluate the quality. So asking them to do so is a pointless exercise (unless if you maybe do it together with them).