FREE REPORT

Is AI Making Us Dumb?

What You Can Do to Stay Sharp in the Age of AI

Based on research by Microsoft Research and Carnegie Mellon University

Published at CHI 2025 Conference on Human Factors in Computing Systems

A companion report for the

Critical Thinking & Problem Solving Skills Course

Something is happening to the way we think

I want to start with a question that bothers me.

When was the last time you sat with a problem, really sat with it, before reaching for ChatGPT or Copilot or whatever AI tool sits in your browser tab?

If the honest answer is "I can't remember," you are not alone.

And that should worry you.

A study published in early 2025 by researchers at Microsoft Research and Carnegie Mellon University surveyed 319 knowledge workers who use generative AI at least weekly.

These workers shared 936 real examples of how they use AI on the job.

The findings confirmed something many of us have sensed but didn't want to admit:

"the more we lean on AI, the less we exercise the mental muscles that make us good at our jobs in the first place."

This report breaks down what the research actually found, why it matters for your career and your life, and what you can do about it.

Because the solution exists, and it is simpler than you think.

What the research actually says

The study, titled "The Impact of Generative AI on Critical Thinking," was presented at the CHI 2025 Conference in Yokohama, Japan.

The researchers surveyed people who work with AI tools regularly, not casual users, but professionals who rely on these tools for their daily work.

Here is what they found, and I think some of these numbers are worth sitting with for a minute.

The researchers identified three specific shifts in how knowledge workers approach their tasks when AI is involved.

None of these shifts are inherently bad, but all of them carry risks if you are not paying attention.

Shift 1: From gathering information to verifying information

Before AI tools, a huge part of knowledge work was finding things out.

You searched databases, read documents, called colleagues, sifted through data.

That process was slow, but it forced you to engage with the material.

You built understanding as a side effect of searching.

Now AI does the gathering.

It delivers answers in seconds.

The new skill is supposed to be verification, checking whether the AI got it right.

But here is the problem:

"verification requires expertise."

You need to know enough about the subject to spot errors.

If you never built that expertise because the AI always did the legwork, you are stuck trusting output you cannot evaluate.

Shift 2: From solving problems to integrating AI responses

People used to solve problems step by step.

Break it down, think through causes, weigh options, pick a path.

That process built mental models of how things work.

With AI, the workflow changes to prompting and refining.

You describe what you need, the AI produces a response, and you adjust the output.

The thinking shifts from "How do I solve this?" to "How do I fix what the AI gave me?"

That is a very different cognitive exercise.

You still need judgment, but you are no longer building the foundational understanding that comes from working through a problem yourself.

Shift 3: From doing work to supervising work

The third shift is from execution to supervision.

Instead of writing the report, you oversee the AI writing it.

Instead of analyzing the data, you check the AI's analysis.

Supervision sounds like a promotion, but it is only useful if you have the skills to catch problems.

A supervisor who has never done the job cannot tell when the job is being done poorly.

The same applies to AI supervision.

If you have not developed deep skills in the work being automated, your supervision is superficial at best.

Why this matters more than you think

You might be thinking: so what?

AI is getting better all the time.

Maybe I do not need to think as hard anymore.

Let the machine handle it.

That argument falls apart for a few reasons.

AI gets things wrong, and confidently

Generative AI does not know what it does not know.

It produces plausible-sounding text regardless of whether the content is accurate.

If you lack the critical thinking skills to evaluate the output, you will confidently submit work with errors you never noticed.

The study found this pattern repeatedly:

"workers who felt underqualified in a subject put disproportionate trust in AI, even when they knew it was prone to mistakes."

Skills atrophy without practice

Your brain works like a muscle.

Cognitive abilities you do not exercise weaken over time.

If you stop solving problems independently, your problem-solving ability declines.

If you stop evaluating arguments, your ability to spot flawed reasoning fades.

This is not speculation.

It is well-established cognitive science.

The researchers warned that over-reliance on AI could leave workers with judgment that is, in their words, "atrophied and unprepared."

Your career depends on what you can do without AI

Here is a thought experiment.

Imagine AI tools disappear tomorrow.

What can you actually do?

If the honest answer is less than you could do two years ago, that gap is growing.

The people who will thrive in an AI-saturated world are not the ones who prompt best.

They are the ones who think best.

AI is a commodity everyone can access.

Your judgment, your ability to evaluate and decide, that is what differentiates you.

The fix: rebuild the skills AI is eroding

The good news is that critical thinking and problem-solving are skills, not talents.

You were not born with a fixed amount.

You can build them, sharpen them, and protect them from atrophy.

But you have to do it deliberately, because your environment is now working against you.

Here is what the research suggests, combined with proven methods for strengthening these abilities.

1. Practice solving problems before you prompt

When you face a challenge, spend at least a few minutes thinking through it on your own before asking AI.

Write down what you think the problem is.

List possible causes.

Sketch out a rough approach.

This does not mean never using AI.

It means using your brain first.

The act of struggling with a problem, even briefly, builds neural pathways that passive consumption never will.

One useful technique is pseudocode:

"writing out your plan in plain language, step by step, before executing."

It forces you to think through gaps, assumptions, and dependencies.

You see problems in your thinking on paper, which is far cheaper than seeing them in your results.

2. Build your ability to spot flawed reasoning

Logical fallacies are everywhere: in news, in meetings, in AI output.

The more you can spot them, the better you evaluate everything around you.

Ad hominem attacks, straw man arguments, correlation-causation mix-ups, slippery slopes, appeals to authority.

These patterns show up in AI-generated content just as often as in human arguments, sometimes more.

When you read an AI response, get in the habit of asking:

a) What assumptions is this making?

b) What evidence supports these claims?

c) Is this actually addressing my question, or is it a well-written non-answer?

3. Learn root cause analysis

One of the biggest traps in problem solving is fixing symptoms instead of causes.

AI tools make this worse because they give you quick surface-level answers.

Root cause analysis is a structured method that asks "why" repeatedly until you get past the symptoms to the real issue.

The five whys technique, for example, takes you from "the car part failed" to "our revenue projections were inaccurate" in five steps.

That depth of analysis is something AI cannot do for you, because it does not understand your specific context the way you do.

4. Develop a growth mindset

The study found that self-confidence is protective against over-reliance on AI.

Workers who believed in their own abilities were more likely to critically evaluate AI output.

A growth mindset, the belief that your abilities improve through effort and practice, is the foundation for this confidence.

This is not motivational fluff.

Research by Carol Dweck at Stanford showed that people with a growth mindset perform better, persist longer, and recover faster from failure.

When you believe you can get better at something, you put in the effort to actually get better.

That effort is exactly what builds the skills AI is threatening to erode.

5. Protect your focus

Critical thinking requires sustained attention, and sustained attention is under siege.

Every notification, every tab, every context switch, chips away at your ability to think deeply.

The irony is that AI tools create more noise to process, not less.

Managing focus is now a prerequisite for thinking clearly enough to use AI well.

Practical techniques include:

time-blocking for deep work, eliminating distractions during

analysis tasks, taking breaks every 45 to 60 minutes, and getting enough sleep.

Your brain uses about 20% of your body's energy.

Treating your physical health as a cognitive performance issue is not optional anymore.

The bottom line

AI is not going away, and it should not.

These are powerful tools that save time and expand what is possible.

But tools that make life easier also make certain skills feel unnecessary.

That feeling is misleading.

The Microsoft and Carnegie Mellon research makes one thing clear:

using AI without strong critical thinking skills is like driving fast without brakes.

It feels great until it doesn't.

The professionals who will succeed in the next decade are the ones who can think independently, evaluate information rigorously, solve problems at their root, and use AI as an amplifier rather than a replacement for their own judgment.

Those skills are not innate.

They are learned.

And they need regular exercise to stay sharp.

You will not just learn about how to turn yourself into a "Decision

Advantage-High Level Leader

You will install it these assets and practice until it becomes second nature.

 Check out The High-Level Leader Thinking System:

Click Here

In a world where AI handles the easy parts, the people who win are

the ones who can handle the hard parts.

The High-Level Leader Thinking System builds the skills that

matters most when the machines are doing everything else.

References

Lee, H-P. et al. (2025). "The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers." Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems (CHI '25), Yokohama, Japan. Microsoft Research and Carnegie Mellon University.

Original publication available at: https://www.microsoft.com/en-us/research/publication/the-impact-of-generative-ai-on-critical-thinking-self-reported-reductions-in-cognitive-effort-and-confidence-effects-from-a-survey-of-knowledge-workers/

Terms and Conditions      Privacy Policy      Anti-Spam Policy      Service Manifesto