In response to "Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task" (Kosmyna et al., 2024)
The recent MIT study warning against "cognitive debt" incurred through LLM use paints a compelling picture of how tools like ChatGPT might dampen neural engagement. But while its conclusions may hold some truth in tightly controlled, neurotypical test scenarios, they don't translate neatly into the real-world contexts of neurodivergent users—especially those with ADHD.
For many of us, LLMs don’t replace thinking. They sustain it.
ADHD and the Myth of Cognitive Laziness
Much of the paper’s argument hinges on EEG data showing reduced neural connectivity in LLM users versus their unaided or search-engine-assisted peers.
This is interpreted as cognitive under-engagement, a kind of intellectual off-ramping. But for those of us managing attention regulation and executive dysfunction daily, this framing is incomplete.
ADHD brains aren't lazy.
They're inefficiently resourced.
Cognitive offloading via LLMs isn’t a shortcut around effort; it's a mechanism to stay engaged longer, to lower the friction of entry, and to scaffold working memory.
The study's authors see AI as a crutch. We see it as an exoskeleton.
Active vs Passive Use: The Missing Variable
The critical variable ignored in this study is intentionality.
The participants assigned to LLM use were often passive recipients, copy-pasting without significant adaptation. That is not how reflective users (especially those in research or coaching) engage with these tools.
In ADHD contexts, LLMs act more like a dialectical partner. We prompt, refine, challenge, reframe. The cognitive load isn’t erased—it’s redirected. Instead of wasting bandwidth on structuring outputs or reprocessing dense academic language, we conserve energy for analysis, synthesis, and critique.
Offloading as Executive Function Strategy
The assumption that cognitive offloading leads to cognitive atrophy assumes a flat baseline of cognitive capacity.
But in ADHD, executive energy is a finite, easily depleted resource.
Offloading doesn’t inhibit growth—it enables it. It allows more of our cognitive resources to be directed toward the why and how of a concept, not just the what.
To label this as cognitive "debt" ignores the reality that many neurodivergent users would be priced out of complex learning tasks altogether without such support.
Risks Worth Naming: Path Dependency and Narrowcasting
That said, there are legitimate concerns here.
Not because LLMs create laziness, but because they can promote cognitive narrowcasting.
ADHD minds are novelty- and clarity-seeking. It's easy to become over-reliant on summarized insights and optimized outputs.
We may become intolerant of ambiguity, less willing to engage with texts that don’t deliver immediate resolution.
This is not decay—it’s a drift. And it can be mitigated with good prompting habits and reflective practice.
Redefining Cognitive Engagement
If cognitive engagement is only defined by internal, unaided effort, then yes, LLMs reduce it. But if we expand the definition to include dialogic thinking, strategic tool use, and sustained inquiry, then the picture changes.
For ADHD thinkers, LLMs don’t replace cognition. They amplify our capacity to engage with it.
So, while the study offers useful caution for educational designers and institutional policy, it misses a crucial nuance: what looks like offloading in neurotypical frames may be survival in neurodivergent ones.
Let’s be careful not to pathologize the tools that are finally allowing us to think out loud, at pace, and in flow.
