AI and the Adult Mind
Key terms
Cognitive offloading — giving work to AI that you used to do yourself is associated with weaker critical thinking.
Cognitive outsourcing — handing over the entire job rather than part of it is associated with atrophy of the underlying skill.
Cognitive laziness — the declining inclination to engage in deep, reflective thinking that develops from doing either one repeatedly.
Cognitive debt — the accumulated cost. You get the finished product now and pay for it in capacity later.
Adults: effects on thinking
When you hand a task to a machine, you skip the effort that would have built the skill.
MIT researchers put people in EEG headsets and had them write essays — some with ChatGPT, some with a search engine, some with nothing.
The brain-only group showed the strongest neural activity. The search group showed less. The ChatGPT group showed the weakest connectivity of all.
Then they took the AI away. The people who had been using it wrote worse essays than people who had never used it at all.
The ChatGPT group also felt less ownership over what they’d written, and their essays were so similar to each other that graders wondered if they’d collaborated.
The researchers named it cognitive debt. You get the finished product now and pay for it in capacity later.
The effect showed up even when participants were told to use the AI as a thinking partner rather than a ghostwriter. Having finished text available suppressed their own reasoning.
This is one small study. It isn’t conclusive, but other research shows similar outcomes.
Outsourcing reasoning is different from outsourcing memory
We’ve always outsourced memory — to notebooks, calendars, other people. What’s different now is outsourcing the reasoning itself.
Heavy AI users score lower on critical thinking
A study of 666 people across different ages and education levels found that heavier AI use tracked with weaker critical thinking, and that cognitive offloading was the mechanism. The author calls the result cognitive laziness: a declining inclination to engage in deep, reflective thinking.
Younger participants offloaded more and scored lower. Higher education was protective.
This is correlational. It doesn’t establish that AI use caused the decline, but the possibility is significant enough to continue to study.
It happens to experts too
Three months after AI diagnostic support was introduced, doctors’ detection rate without it had dropped from 28% to 22%.
A year-long study of oncologists found that AI raised their accuracy while their confidence in their own unaided judgment declined.
These are people at the top of their fields, losing skill they’d spent decades building.
Thinking deficits can become emotional harm
Loss of the ability to trust in one’s own discretion is a major concern for Adeptable Mind Coaching. When people can no longer trust their own judgment because of reliance on AI, it creates a slippery slope for mental health across the board. We can expect this to lead to lower self esteem, higher depression, and related psychological side effects.
Effort matters
Effortful processing is what builds durable skill. When you struggle through something, the struggle is a necessary piece of what produces the learning.
And although AI can genuinely help you complete a task, it also removes the exercise that would have made you better at the skill.
So before delegating a task, especially if it’s going to be repeatedly doled out to AI, be aware that the cognitive skills that were required for you to have accomplished that job by yourself will likely atrophy, unless you find another application that requires the same skill that you continue to do yourself, so that you ensure that cognitive capacity stays sharp.
Heavy media multitasking has also been associated with lower gray matter density in the anterior cingulate cortex, a region used for focus and emotional control. That study was small, seventy-five adults, and looked at people at a single point in time, so it can’t say which way the arrow points. Heavy multitasking may be wearing the region down, or people who start out with less there may be more drawn to multitasking in the first place. Nobody has run the long-term study that would settle it. If it turns out to be the first, it matters a great deal, and it shouldn’t be set aside while we wait.
Before you read on, notice how long you’ve been sitting still. Your body might be asking you to move.
Adults: effects on emotional wellbeing
Relationships
A twelve-month study of more than 2,000 adults across four countries found that increased social chatbot use predicted increased loneliness. Not the reverse — the chatbot use came first.
Stanford researchers studying Character.AI users found that people with limited offline social networks felt more lonely after seeking emotional support from chatbots. The people who most needed connection were made worse by the substitute.
Scrolling and depression
Endless short-form content drains your mental reserves, which produces burnout. Burnout raises stress and anxiety. Sustained stress and anxiety produce depression.
The researchers describe the endpoint as a defensive shutdown — what’s left when the resources for coping have been spent.
Older adults
A meta-analysis covering more than 400,000 adults over fifty found that technology use did not contribute to cognitive decline in that group, and was associated with reduced risk. The digital dementia hypothesis, for that generation, isn’t supported.
The likeliest explanation is how they use it: email, video calls, looking things up. It’s active use.
Exceptions
Research cited by the American Psychological Association found that people on the autism spectrum can use AI as a lower-stakes place to practice social skills: no eye contact, no unpredictable cues, no cost to getting it wrong.
Even then, the practice has to transfer. A flight simulator is useful because the pilot eventually flies in real weather. Real people are far less predictable than any chatbot, so the skills practiced there need a ramp to the world the person actually lives in. Without that ramp, what started as an advantage can become a handicap.
Additionally, the risks in that population are specific. Dependency can amplify social withdrawal rather than easing it, and there are documented cases of chatbots encouraging harmful behavior or failing someone in crisis.
There’s also a tension around masking. With the chatbot itself, autistic users describe relief — a place to communicate without performing. The risk comes from using AI to mask with other people, rewriting communication to sound neurotypical. Masking normally requires reading the room; outsourcing it means that skill doesn’t get exercised. And it creates a mismatch — what someone sees in your email isn’t who they meet in person.
The same sycophancy that makes these tools comfortable has also been documented reinforcing delusional thinking in some autistic users.
Taking care of your own mind is the point of all this. Step away for a while if you need to; the page will be here.
Helpful versus harmful
The key distinction in this research is active versus passive use.
Active: writing the first draft yourself, searching and evaluating what you find, managing a feed that takes analysis, reflection, and judgment. Using AI to argue with, then checking what it told you is another example.
A live conversation by phone or video is more demanding than texting. You’re processing tone, timing, and someone’s reaction in real time, and you can’t edit before you speak. Texting isn’t as harmful as scrolling, but it isn’t as beneficial as talking.
Passive: scrolling a feed, binge watching, letting the next thing play, accepting an AI’s answer without testing it. Testing means checking the sources yourself, not asking a different AI whether the first one was right. These are the kinds of things that will get you in trouble. The same goes for reading a meme and reposting it without checking the claim.
A review of seventeen studies of adults over forty found active screen use associated with better memory, executive function, and attention. Passive use was associated with declines in verbal memory and overall cognition.
Passive use is measurably harmful.
Research in education found the same split. Guided, reflective use of AI supports critical thinking. Unstructured or heavy reliance shifts the work away from analysis, toward less critical evaluation of what the AI produced.
Being active isn’t enough
Active use protects your cognitive function better than passive use. It does nothing to protect you from the manipulation itself.
The upshot
Nobody knows the long-term picture. These tools are too new and the studies are too young.
What we do have is a consistent direction: you lose capacities when you don’t use them. Skills you stop exercising get weaker. Reasoning you outsource atrophies. And the feeling of connection we can simulate with AI doesn’t feed us the way human connection does. We need the biochemical lift that comes from being with another being with the capacity to connect and empathize.
If we don’t want to lose the mental capabilities we currently possess, or at least want to minimize what’s getting lost, we have to be deliberate in a way most of us currently aren’t.
Get in the habit of asking yourself — am I engaging in passive use or active — ideally before you start. Catch yourself in the middle of something passive like scrolling and shift to active use, or simply stop and go do something in real life. Remember that everything on social media and every AI chatbot is optimized to hold you there. That’s attention capture.
Noticing when you’ve stopped thinking and started accepting what’s being said by an AI or a meme or anyone online is a critical step out of their manipulative design, even if you don’t change what you do right away.
This is coaching, not therapy. I don’t diagnose or treat. This work is not a substitute for medical or psychological care.