Is ChatGPT Dumbing Down your Kid? New MIT Study Says, “Yes.”

Is ChatGPT dumbing down your kid? It is and here’s what you can do.

A new MIT study reveals the powerful consequences of artificial intelligence on actual intelligence, and guess what? Simply (and terrifyingly) put, the use of artificial intelligence undermines your child’s actual intelligence. In short, when children don’t think for themselves, they don’t learn to think for themselves. That should surprise no one.

I’ll get to the disturbing details of the study in a moment, but let me first explain why these outcomes were obvious and inevitable. In a nutshell, the brain functions like a muscle insofar that it becomes stronger when it is used and atrophies when it is not used. I could list a thousand additional factors that affect thinking, but that simple premise really is enough for this discussion.

And when I say that the brain functions like a muscle, most people think I’m speaking overly metaphorically. I’m not. While the brain, of course, isn’t actual muscle tissue, its functioning is remarkably similar. Much in the way that exercising muscles builds more muscles, exercising the brain builds the brain—literally. Every single time we engage in a thinking act, the brain builds more wiring, such as synapses through synaptogenesis, for that thinking act. On the flipside, the brain not only allows existing pathways to diminish when they’re not used, it actually overwrites existing pathways with new ones.

Watch this play out in the MIT study …

The MIT Study

That study is Your brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task, by a team of researchers led by Dr. Nataliya Kosmyna. The scientists broke a group of students down into three essay-writing groups: An “A.I.-assisted” writing group that used multiple LLMs (not just ChatGPT), a “search engine” group, and a “brain-only” group. The students then engaged in three writing sessions while the researchers monitored their brain activity using an EEG. Each student was interviewed after each session, and all of their writing was assessed by humans, as well as an A.I.

So, what happens when one group is required to use their brains more than the other groups? Would it shock you to know that the group that needed to do their own thinking actually thought more? I hope not, not anymore than it should be surprising that a group of kids who practiced hitting a ball did better at hitting a ball than a group of kids who watched a robot hit a ball for them. (Okay, that’s not a perfectly fair analogy to the A.I. usage in this case, but it illustrates the point.)

And the point is that brain-only group performed better and scored higher on their essays. But that’s not the most important outcome for us. What’s more important is that “the brain-only group exhibited the strongest, widest-ranging networks” of brain activity, while the group with A.I. “assistance elicited the weakest overall coupling.” In other words, the brain-only group thought a lot; the A.I.-assisted group did not. Do you remember what we said about what happens when the brain “muscle” isn’t used?

But it gets worse. The researchers brought those two groups back for a fourth session and switched their roles. They gave the A.I. group a brain-only writing task and the brain-only group an A.I. writing task. And here’s what’s so important: the brain-only group still performed better, even when using A.I., and the A.I. group still performed worse, even when given the opportunity to think for themselves. Or should I say, it did worse because they now had to think for themselves.

Over the first three brain-only writing assignments, the brain-only students built their brains for the task, and they built mental frameworks (read: habits) to rely on when engaging those tasks. Thus, that they then “gained” an A.I. assistant did not suddenly degrade all of the wiring that their brains built. But the A.I. group, when suddenly given the opportunity for a brain-only task, not only had built no wiring for accomplishing that task, it also, and this is the most critical part, created wiring and mental frameworks for using A.I. instead.

What that means in a nutshell, and these are my words not those of the study, is that the brain-only group got smarter and the A.I. group not only failed to become smarter, they got dumbed down—they became habituated to relying on A.I. Thus, when given the opportunity to do so, they were incapable of thinking as well as the brain-only participants did.

All of that should be concerning enough, but there’s more. In addition to the direct cognitive effects, the researchers also found that brain-only participants “demonstrated higher memory recall” and engagement of thinking-related brain areas compared to the A.I group. Meanwhile, compared to the brain-only group, the A.I. participants reported lower “ownership of their essay,” which is an educator’s way of saying that they didn’t care about it as much and did not feel as though it was their own.

Thus, to sum it all up, A.I.-assisted writing made the kids perform poorly, made them dumber, and made them less invested in their own thinking and writing.

What to do

In light of this study, one school of “thought” could be that since everyone is going to rely on A.I. in the future anyway, kids who do so will be no worse off than their peers, and using A.I. might free up time for them to do things that are more valuable than writing essays, which, again, they won’t really ever need to write on their own anyway because A.I. will be there to “assist.” Those who subscribe to that position probably should stop following me here at Actual Intelligence right now as we will be rather inclined to disagree.

The other school of thought is that thinking skills, such as those developed through writing, which research repeatedly shows is the best way to teach critical thinking, are far more important than any and all expediencies achieved through A.I. assistance. Let me rephrase that: If you want your kids to build their brains rather than have them degenerate into relatively useless gelatin that can only write A.I. prompts or order burrito online, then keep their brains as far from A.I. as possible.

Obviously, there’s not much that you can do with your college-aged kids other than share this information with them and hope they make the right decisions. But for kids still under your roof, there are things you can do:

1. Share this information with them. Most kids don’t want to become dumber; they do value their ability to think. So, take time to explain, and then reinforce, the consequences of A.I. In fact, start thinking of A.I. as something about which you need to begin messaging no differently than alcohol, drugs, and sex.

2. Ask them how they use A.I. Understand their current relationship with A.I., and please keep in mind that the MIT study does not speak to other ways that students might interact with A.I. beyond this one context. Using A.I. in other ways might be more or less consequential.

3. Check their work: There are plenty of sites out there that scan essays to see if they were written by A.I. Those sites are not perfectly reliable, but they might offer useful information about what your kid is up to.

4. If you want to get serious, have your kids download all their source materials before writing, then shut of their internet while they write. Take away the temptation; make them use their brains.

Conclusion

The implications of A.I.-based “thinking” work are becoming clear, but for anyone who has thought about it or who values thinking, they’re also not surprising. Every time we use A.I. to “assist” our thinking, it not only prevents us from thinking, it degrades our capacity to think in the future.

Worse—much, much worse—is that those of you reading this built your brains before A.I. existed, which means that even if you gravitate to using A.I. now (please don’t), you’ve got a lot of “muscle” built up to abate its consequences. A.I. will still degrade your thinking, but those sound neural pathways you built up all your life won’t all turn to jelly overnight.

But for your kids, it’s different. Their neural pathways are still in the process of building up for the first time. Even though we are all always rewriting our brains, kids’ brains have not even fully developed, so whatever they habituate to will become hardwired moving forward. Consequently, kids who are raised as A.I. natives might never develop their brains for thinking in the same way yours did. And that will not only affect their lives, but a generation of lesser-thinkers will affect all our lives.

But there’s good news! Somewhere down the line, kids who actually learn to think for themselves will stand out against the emerging generation who might not. So, if you can raise your own child to think critically, they might just be among the few who lead the world to a better place.

And that, once again, is why actual intelligence is so important.

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit pearlmanactualintelligence.substack.com