📉 AI May Be Dulling Mental Effort, Suggests Groundbreaking EEG-Based Research
Cambridge, MA — June 2025 — A new study from the MIT Media Lab has raised red flags over the cognitive effects of generative AI tools like ChatGPT. Researchers found that users relying on AI assistance for writing tasks exhibited significantly lower brain activity, weaker memory retention, and a decline in creative quality over time.
The early-stage research, led by Dr. Natalia Cosmina, is among the first to quantify how AI affects mental engagement using neurological data.
🧪 How the Study Was Conducted
The study involved 54 participants aged 18–39, who were asked to write SAT-style essays under three different conditions:
- ✍️ Brain Only – No tools or internet access
- 🔍 Google Only – Allowed to search online
- 🤖 ChatGPT Assistance – Prompted and edited using AI
Each participant wore a 32-channel EEG cap that monitored their brain activity across key frequency bands: alpha, theta, and delta.
📊 Key Findings
🧠 Brain Activity
- Participants using ChatGPT showed the lowest neural engagement across all frequency bands.
- The “brain only” group exhibited the highest sustained mental effort.
- When GPT users had to write again without AI, their performance dropped sharply.
🧾 Essay Quality
- Essays written with ChatGPT assistance became more uniform and “soulless” by the third attempt.
- Participants began copying and pasting full AI-generated prompts with minimal editing.
- Human raters found the writing to be lacking in originality and depth.
🧠 Memory & Retention
- On a surprise recall test:
- Brain-only writers remembered their essay content far better.
- GPT users struggled to recall key points they had “written.”
- The brain-only group also reported higher personal satisfaction with their work.
🔁 AI Revision vs. AI Generation
One unexpected twist:
- When brain-only writers used AI to revise their drafts, their EEG activity spiked, indicating increased cognitive stimulation.
- This suggests that AI may be more beneficial as a revision tool than a first-draft generator.
🔎 Implication: The most effective workflow may be to write first using your own cognition, then revise with AI assistance.
⚠️ Ethical and Educational Concerns
Lead researcher Natalia Cosmina says the study was fast-tracked to warn educators and policymakers:
“We’re seeing signs that over-reliance on AI tools could lead to cognitive offloading, where users essentially stop thinking.”
She also embedded “hallucination traps” in the study — statements designed to trick AI into inventing false citations (like claiming “GPT-4 was used” when it wasn’t), which ChatGPT did multiple times.
🔍 What’s Next?
MIT is now conducting a follow-up study focused on software developers using AI autocomplete tools (e.g., Copilot). Early data suggests even more pronounced cognitive erosion, especially in code understanding and recall.
📌 Final Thoughts
This week’s AI landscape presents a complex picture:
- OpenAI is releasing powerful tools like its multi-agent SDK.
- YouTube is scaling AI-generated Shorts with Google’s V3 model.
- But institutions like MIT are reminding us of the human cost of convenience.
As AI becomes more embedded in daily life, thoughtful adoption—not just adoption—is key.