Illusions Of Ai

You can relate

I am sure you will keep nodding your head in agreement as you read this! Have you ever used ChatGPT or Perplexity AI (or the rest of them) to sharply do an email, post online, or one schoolwork like that? You felt brilliant about the result, only to realise an hour later that you could barely remember what you wrote.. Yes, no be only you! A new preliminary study from a team at MIT titled “Your Brain on ChatGPT” is digging into this exact feeling, and its findings are a big deal for anyone who cares about real learning. (Read the actual paper here)

The paper introduces a powerful idea: cognitive debt. It’s the price we pay when we outsource our thinking to AI without first doing the mental gídígbò (heavy lifting) ourselves. For educators, parents, and learners across Africa, where we’re rapidly adopting these new tools, understanding this concept isn’t just academic—it’s essential.

Why Was This Study Even Done?

As an educator, my core mission is to help develop the human brain—to make people get it and develop intellectual independence. But we’ve entered an age where the traditional cycle of learning has been disrupted. The old model, where a teacher determines if a student has learned based on written submissions or take-home evaluations, is no longer effective. Students now have access to a source of knowledge as vast as the teacher’s and can produce a flawless essay with a few prompts, making it impossible to reliably measure their true grasp of a concept by the final output alone.

The real concern isn’t about catching cheaters (unfortunately, there will always be back-benchers); it’s about what happens to the brain when it’s no longer required to struggle. The researchers at MIT wanted to know the idi abájo (look under the hood). They focused on concepts that are the bedrock of education:

• Cognitive Effort: The mental gídígbò that builds deep understanding and memory.

• Critical Thinking: The ability to analyse, question, and form original thoughts.

• Intellectual Independence: The power to think for yourself, without digital support.

The study was born out of an urgent need to understand if we are trading these vital skills for the convenience of AI-generated answers.

How Did They Figure This Out?

The MIT team designed a clever experiment. They took 54 students and split them into three groups for an essay-writing task:

1. The LLM Group: Used ChatGPT for assistance (a.k.a the AI Group).

2. The Search Engine Group: Could only use a search engine for research.

3. The Brain-Only Group: Had no access to any technology.

But here’s the most fascinating part: while the students worked, they wore  Electroencephalography (EEG) caps that measured their brain activity. Think of it like a master chef watching a team of cooks in a kitchen. They weren’t just tasting the final meal (the essay); they were observing the entire process— especially how different parts of the brain (the cooks) communicated, collaborated, and engaged to create the dish.

In the fourth session, only 18 participants returned, and they were reassigned to the opposite group. So, LLM users went “LLM-to-Brain” (no tools), and Brain-only users went “Brain-to-LLM” (using ChatGPT). They wrote on topics they had already tackled in previous sessions. This switch allowed researchers to observe how previous tool use affected their brains when the tools were removed or introduced. This switch revealed everything.

What Was Discovered?

The first three sessions repeated this same format. This established a clear pattern, with the first session capturing the most natural responses before participants knew what to expect. By the second and third sessions, the patterns held, but participants were more familiar with the task.

The fourth session, however, took them by surprise with an unexpected twist. Only 18 participants returned, and they were reassigned to the opposite group. So, AI users went “LLM-to-Brain” (no tools), and Brain-only users went “Brain-to-LLM” (using ChatGPT). This switch is why the most powerful insights are drawn from comparing the first and fourth sessions. The findings, backed by stark numbers, were staggering and revealed four key truths:

While you use AI, Your Brain is Quiet—Too Quiet.

The scanner the volunteers had on their heads while writing the essays showed that using ChatGPT dramatically reduced brain connectivity. The AI-only group demonstrated up to 55% weaker brain connectivity compared to the Brain-Only group. It makes sense that the LLM group’s brains were significantly quieter because the AI was offloading the mental work, as one researcher put it, there was “much less chatter happening” between different parts of the brain. While the Brain-Only group’s minds were busy cooking up ideas, combining and connecting concepts.

Very little Recall and Lack of Ownership.

This mental graveyard had a direct, measurable impact. A shocking 83% of ChatGPT users couldn’t accurately quote from their own essays just minutes after writing them. In stark contrast, only 11% of the Brain-Only group and 16.7% of the Search Engine group had the same issue.

The AI users also reported a “fragmented sense of ownership,” feeling disconnected from the words on the page because they hadn’t wrestled with the content to make it their own. Obviously!

He never finish o, as if that wasn’t enough, no one, not even one person in the AI group could correctly quote from the essay they had just written! If that does not scare or alarm you, I don’t know what will.

The “ChatGPT Vibe”: Originality Gets Replaced by Homogeneity.

The essays from the AI group were described by human graders as “soulless” and statistically similar to one another, which gave off that “ChatGPT-ish vibe.” They had good structure and language but lacked personal insight and originality. At this point eh, we have all gotten familiar with how ChatGPT sounds; this is no surprise because of the underlying predictive architecture behind the chatbot. This points to an “echo chamber effect,” where AI reinforces existing ideas instead of fostering diverse, creative thought.

The Session 4 Twist: Timing is Everything.

This session shouted loud and clear: WHEN you bring AI into the learning process matters immensely.

The folks who relied on ChatGPT initially struggled big time when they had to go solo. A whopping 7 out of 9 (78%) couldn’t even quote from their own essays, and only 1 out of 9 could get a quote right. It was like their brain hit Olumo rock, Omo!

However, when the Brain-Only group was given AI, 78% of them could quote their work well. The participants who started “brain-only” and then got ChatGPT were able to maintain “very good ownership” of their work, recalling and summarising accurately. It seems building those mental muscles first is the koko to real, lasting understanding.

The conclusion is clear: over-reliance on AI from the start impairs memory, kills originality, and creates a shallow “illusion of learning.”

Who Does This Affect?

Students and Lifelong Learners

Our brains are designed to be efficient; they will always look for a shortcut. If we constantly delegate the struggle of thinking to AI, we deny our brains the very exercise they need to grow. Learning happens during the wrestling, not in the final, clean output. Relying on a tool to think for you is like taking a pill to induce sleep instead of addressing the stress that keeps you awake. Eventually, your body forgets how to sleep on its own. Our brains are no different.

Professionals and Creatives

For those of us in fields that require problem-solving, strategy, and innovation, oya, come closer, this is a serious warning. If we offload the entire creative process to AI, we gradually lose our ability to spot patterns, connect ideas, and generate true insights. What happens when the internet connection shows you shege or you’re in a boardroom and need to think on your feet? We must cultivate the intelligence we have (I will get to that in a bit), not just the one we borrow from AI.

So, Kini Big Deal? Here’s How We Use AI Without Losing Our Heads.

The problem isn’t the technology itself; it’s the timing and the intention of its use. So, what’s the way forward for us as educators, students, and professionals in the AI age? Let’s reason together:

For Educators & Trainers

We must urgently re-evaluate how we assess learning. The old model, where a teacher determines a student’s knowledge based on written submissions or homework, is no longer effective. Students now have access to a source of knowledge that is as great, if not greater, than the curriculum and can conjure a polished answer with a few prompts. The era of grading submitted assignments alone is over (Ọpọ́n ti sún). We need to shift focus from the outcome to the process. This could mean more in-class writing, oral defences of projects, or marking the steps and thought process, just like we do in mathematics. The goal is to design tasks that minimise the temptation to outsource thinking.

For Students & Lifelong Learners

Adopt a “Brain First, AI Second” approach. When I tackled this 206-page research paper, I intentionally did the rigorous work first—I read it, studied the graphs, and formed my own thoughts. Only then did I use AI to brainstorm with me and critique my ideas.

• Struggle with the material first. Let your brain wrestle with the concepts.

• Form your own initial thoughts, questions, and outlines.

• Then, use AI as an intelligent partner. Ask it to test your understanding, explain a concept differently, or play devil’s advocate to your argument.

• Alternate your methods. Sometimes, just put the tech away and see what your own mind can produce (scary, but much needed if you want to be relevant)

For Professionals & Creatives

The danger is real. If we let AI do all the heavy lifting, we risk losing the very craft we’ve built—our ability to see patterns, connect ideas, and create something new. The way forward is to treat AI as an intelligent partner, not a replacement. Do the rigorous work first. Form your own strategy. Wrestle with the problem. Then, bring in AI to critique your ideas and help you see what you might have missed. Use it to amplify your thinking, not to outsource it. Our natural intelligence must lead; artificial intelligence should follow. This way, AI becomes an Augmented Intelligence and not artificial!

For Everyone in between

This whole thing boils down to a simple, timeless principle: Use It or Lose It. While you won’t literally lose your brain, you risk letting it get mentally lazy. If you don’t exercise a muscle, it gets weak. The same goes for our cognitive skills. We must consciously resist the urge to outsource every minor mental task. I’ll be honest, the other day I was about to send a prayer message, and my first instinct was almost to ask Gemini to help me pray! That’s how deep this can get if we’re not careful. We stop developing our ability to think, to create, to even connect spiritually, because we have a tool that offers an easy way out. The real danger isn’t that AI will take over, but that we’ll willingly hand over the keys to our own minds.

Here’s the mindset shift we all need to make: AI should make us better at our work, not just make better work for us. There’s a huge difference. When AI just produces the final essay, report, or email for you, it has done the work, and you are left with a good-looking product but no personal growth. But when you use it to challenge your own ideas, to learn a concept more deeply, or to handle the repetitive tasks so you can focus on the core strategic thinking—that’s when it makes you better. The goal isn’t to have a perfect document; the goal is to become the kind of person who can produce that level of work, with or without the tool. AI should be the gym equipment, not the personal trainer who lifts the weights for you.

The big deal is this: we cannot afford to raise a generation that is dependent on machines to think for them. True learning is a biological process, not a digital one. Let’s use AI to amplify our intelligence, not replace it.

P.S. My team at Kini AI is currently reproducing this research in Nigeria, focusing precisely on the effect of AI on students and teachers in the pedagogical process. If you want to be a part of this important work, reach out to us at info@kini-ai.com.