Against Brain Damage
Source: One Useful Thing
Author: Ethan Mollick
Original source: https://www.oneusefulthing.org/p/against-brain-damage
Source type: essay
Private backup: the full article text is archived in the private repository at archives/articles/oneusefulthing-org-against-brain-damage.source.md. It is not published on the public Quartz site.
Summary
Mollick rejects the sensational claim that AI causes literal “brain damage,” but argues that careless AI use can weaken habits of thinking by outsourcing effort that matters for learning, creativity, and collaboration. He reviews evidence suggesting that unguided ChatGPT use can reduce student learning, while guided, well-prompted AI tutoring can improve outcomes substantially. For creativity and writing, he warns that AI can anchor users to its suggestions and reduce ownership, so people should think, draft, and generate ideas before asking AI to expand or critique them. For teams, AI should improve meetings and collective intelligence rather than replace human interaction with summaries and avatars.
Pull quotes
Not brain damage, but thinking habits
“AI doesn’t damage our brains, but unthinking use can damage our thinking.”
— Mollick, distinguishing literal brain-harm claims from habits-of-mind concerns.
The default AI shortcut
“The problem is that even honest attempts to use AI for help can backfire because the default mode of AI is to do the work for you, not with you.”
— Mollick, on why unguided AI help can undermine learning.
Sequencing creativity
“Always generate your own ideas before turning to AI.”
— Mollick, on avoiding AI anchoring during idea generation.
Think first
“Just remember to do the thinking first, because that’s the part that can’t be outsourced.”
— Mollick, on using AI after the human thinking work has begun.
Big ideas
- Learning still needs some struggle, even when AI can make things easier
- AI simulations need clear boundaries for learning
- Students need to bring the purpose; AI should not supply it for them
- Education should teach thinking with, without, and about AI
Claims
- AI can undermine learning when students use it without guidance
- Learning requires some productive struggle that AI can remove
Key evidence and examples
- The MIT Media Lab “Your Brain on ChatGPT” study is treated as evidence of reduced engagement and memory, not literal brain damage.
- A Turkey high school study found unguided GPT-4 homework help associated with worse final exam performance.
- Studies from Nigeria, Harvard physics, Stanford programming, and Malaysia are cited as evidence that guided AI use can produce learning benefits.
- Mollick’s practical rule is to think, draft, or meet first, then use AI to expand, critique, or improve the work.
Education relevance
Very relevant for AI pedagogy, tutoring design, writing instruction, academic integrity, prompt design, and teacher guidance because it distinguishes productive AI scaffolding from cognitive outsourcing.
Durability note
The particular studies will age as the evidence base grows, but the sequencing rule—think first, then use AI for critique or expansion—is a durable classroom and knowledge-work heuristic.