A study from MIT Media Lab Research, “Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task,” asked fifty-four participants to write essays, with some participants using their own heads, some using a search engine, and some using a large language model (LLM). The conclusion:
While LLMs offer immediate convenience, our findings highlight potential cognitive costs. Over four months, LLM users consistently underperformed at neural, linguistic, and behavioral levels. These results raise concerns about the long-term educational implications of LLM reliance and underscore the need for deeper inquiry into AI’s role in learning.Everywhere in academic life, teachers are hearing the mantra “Our students need to know this.” I’d offer great resistance. What students first need to know is how to read well and how think, speak, and write clearly and persuasively and ethically. (I swiped part of that sentence from a comment I wrote a couple of days ago.)
A detail from the abstract that I found interesting: “LLM users also struggled to accurately quote their own work.” That reminded me of a low-tech strategy teachers sometimes use to help figure out whether a student has plagiarized: present the student with a passage from their writing with every fifth word removed. It’s surprisingly difficult to recreate what isn’t your own.
[I know: just fifty-four participants. But the findings were consistent.]
comments: 3
Not only do we want them to learn such skills, but we want them to do them to act on their skills significantly better than folks who didn't go to post secondary, or else there's no point.
In pop culture terms, we want them to be augmented in such skills, not augmented in knowledge by having an information plug-in in their heads.
I think that knowing how pretty much implies doing — I don’t know what it would mean to know how to read well, etc., without doing so. But I take your point that doing so requires effort, which the use of AI discounts — though you still have to be able to write clear prompts.
I guess I was trying to say that if they don't "know" enough to "do," then we are kidding ourselves that they are learning in the first place.
I'm thinking of student protesters coming to Banff for the 2002 G-7, who couldn't explain to the public here why the G-7 was bad. Unless all they wanted to do was somehow, through pathos without logos, reach the seven leaders, they might as well have stayed home. Incidentally, the local chief of police was surprised at the lack 2025 protesters; he speculated it was because you can now digitally protest without traveling.
Post a Comment