MIT study finds impact on brain activity and memory in essays written with ChatGPT, Google, or no tech.

“Humanity has always been great at building tools. But when ChatGPT was released and people started reporting memory issues, we thought…hmm, interesting.”

Nataliya Kosmyna, PhD, researcher at MIT Media Lab

AI-generated image with Gemini

Hi everyone, have a great week ahead!

By José Brito, journalist and founder of Pupa Educação Digital.

Every week, as I sit down to write the next Radar Pupa article, I search for that spark—a quote that aligns with the spirit of our time. A phrase that bridges media, education, and technology in a way that actually reflects reality. And let’s be honest: keeping up with today’s world isn’t for amateurs. Everything is fleeting. Sometimes hostile, shallow, or artificial. Yet also dazzling, useful, and fast. It rhymes because it’s true.

This week, I want to talk about writing. About the essential skills that go into crafting a solid essay—presenting a theme, building a perspective, supporting your views with arguments and data. But also, writing in a voice that’s uniquely yours. That human touch.

But here’s the catch: how do you write about something you’ve never really engaged with? How do you argue a point you didn’t research, didn’t prepare for, didn’t fully think through? What happens when someone more informed enters the conversation and challenges you—with evidence, lived experience, or just plain brainpower?

That’s the kind of question being asked in a recent experiment led by Nataliya Kosmyna,

at the MIT Media Lab, featured in Brazil’s Fantástico TV program this Sunday. Kosmyna gathered 54 students from around the world, aged 18 to 39, and gave them 20 minutes to write an essay. Participants were split into three groups:

• Group 1 could use ChatGPT.

• Group 2 used Google for research.

• Group 3 used only their brains—pen and paper (or keyboard, if preferred).

The essays were written, and then the real test began. Kosmyna used real-time brain-monitoring equipment to analyze how students’ brains worked during the process—and afterward, what they actually remembered.

The findings were striking:]

• A whopping 83% of students in Group 1 (the AI group) couldn’t recall a single phrase from their own essay.
• Groups 2 and 3 showed significantly higher brain activity, with Group 3—the no-tech group—performing best.

The takeaway? We still have a long journey ahead when it comes to integrating AI into education and the workplace in a meaningful, balanced, and human-centered way. It’s a constant effort. Trial and error. And it calls for critical thinking, diversity of perspective, and thoughtful content moderation. We are now entering the second wave of AI adoption in schools. According to Brazil’s 2024 ICT Education survey, 70% of high school students have already used AI tools for schoolwork—yet only 32% did so under teacher supervision.

Let’s move forward. I’ve shared the full MIT study link below: “Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task”. And while we’re on the topic of writing, don’t miss this week’s tip from Sarah Schollmeier, a TikToker I absolutely love, who just dropped a killer breakdown of how to crush the ENEM essay this November.

#01 Brain activity stimulation becomes the focus of new studies on AI in schools

#02 Brazil moves forward with legislation to protect children and teens online

#03 AI experimentation drives Brazil’s market and sparks cross-department collaboration in companies

#04 National AI Olympiad opens doors to international opportunities and scholarships

#05 Educa Week to be held October 8–10 in São Paulo with fresh debates on education and technology

📚 READING TIP

The scene is familiar. It’s evening. Dinner is almost ready. A 12-year-old is getting ready for a shower, notifications buzz on your phone, and you’re trying to center yourself for the week ahead.

Suddenly: “Alexa, stop!”
My daughter has had enough of the voice assistant reading the weather forecast. Meanwhile, I’m trying to soak in every word. “Tomorrow in Rio de Janeiro: clear skies, highs of 28ºC, possible rain late afternoon.” Love it. But apparently, I’m the only one who does.

Flash forward to the next morning—coffee brewing, butter on toast, microwave beeping. Then comes the tablet notification. Tum-tum. Time to take my kid to school. We exchange a glance, the same one as the night before. A small, knowing smirk. It’s an inside joke now. It happens every day. A reminder of how we let technology into our private lives—how we rely on it, criticize it, and still ask it to stay.

This week’s reading tip comes from that very place of tension and reflection. It’s a book I read last year when I started my postgrad in AI for Business at IBMEC. In How Humans Judge Machines, Cesar Hidalgo flips the narrative: we’re not just judging the machines—they’re reflecting back how we judge ourselves.

A must-read for anyone navigating this increasingly blended world of humans + AI.

» How Humans Judge Machines

Author: César Hidalgo
Publisher: MIT Press
Year: 2021

Catch you next week! 🙂

Share

Leave a comment