Skip to content
Go back

Critical Thinking: Intelligence Doesn't Protect You (the Data Says So)

Illustration of a brain with a magnifying glass examining connected ideas

Critical thinking sits near the top of almost every in-demand skills ranking, usually described as a general capacity: you learn it once and then apply it anywhere. It’s a convenient description, and it’s largely wrong.

The research on this has two results worth holding together, because on their own each one is misleading. First: measured intelligence doesn’t protect you from the most common reasoning errors. Second: critical thinking can be taught, but it works far better inside a specific domain than as an abstract course.

The test almost everyone gets wrong

In 2005 Shane Frederick published a three-question test in the Journal of Economic Perspectives, the Cognitive Reflection Test. The first question is this:

A bat and a ball cost $1.10 in total. The bat costs $1.00 more than the ball. How much does the ball cost?

The answer that arrives first is ten cents. It’s wrong: if the ball cost ten cents, the bat would cost $1.10 and the total would be $1.20. The right answer is five cents.

Frederick administered the test to thousands of American university students. Even at the most selective universities, the majority got at least one of the three questions wrong. At MIT the average was around 2.18 correct out of 3; at other institutions it fell below 1.

The point isn’t that people can’t do arithmetic. The arithmetic is trivial. The point is that there’s an answer that arrives before the correct one, and that the work of critical thinking consists almost entirely of not stopping there.

Intelligence doesn’t protect you

Here comes the most uncomfortable result. Keith Stanovich and Richard West published a study in the Journal of Personality and Social Psychology in 2008, on a large sample, comparing measured cognitive ability with susceptibility to a range of biases.

On some biases, intelligence helped. But on myside bias, the tendency to evaluate arguments and evidence more harshly when they contradict your own position, the correlation with cognitive ability was essentially zero. People with high cognitive scores weren’t less biased. They were just better at building arguments for the side they were already on.

Dan Kahan and colleagues found something even sharper in a 2012 study in Nature Climate Change. In a representative sample of US adults, those with higher science literacy and numeracy were not less polarised on climate risk. They were more polarised: technical skills were used to reinforce their cultural group’s position, not to correct it.

Cognitive skills aren’t neutral with respect to motivation. If you want to reach a particular conclusion, being clever helps you get there.

Why it isn’t a generic skill

Daniel Willingham, a cognitive psychologist at the University of Virginia, wrote a widely cited 2007 essay in education whose title is already a thesis: why is critical thinking so hard to teach.

The argument is that what we call critical thinking isn’t a procedure separable from content, but depends massively on how much background knowledge you have about that specific content. Recognising that a figure is suspicious requires knowing what values are plausible in that field. Recognising a weak analogy requires knowing both sides of the analogy.

André Tricot and John Sweller stated the strongest version of this position in a 2014 article in Educational Psychology Review: generic problem-solving skills can’t be taught because we already possess them, and what distinguishes the expert from the novice is specific knowledge organised in long-term memory, not a better reasoning technique.

The practical consequence: a critical thinking course disconnected from your work almost certainly produces less effect than six months spent discussing the real problems of your field in a structured way.

But teaching it does work, if you do it a certain way

That said, the literature doesn’t say it’s pointless to teach. It says how you teach it matters a lot.

In 2015 Philip Abrami and colleagues published a meta-analysis in Review of Educational Research collecting 341 effect sizes from studies on teaching critical thinking. The overall average effect was positive and moderate (g ≈ 0.30). But the useful part is the moderators:

Teaching elementEffect on effectiveness
Structured dialogue (discussion, debate)increases
Authentic, real-world problemsincreases
Mentoring or expert supervisionincreases
All three togetherlargest effect in the meta-analysis
Critical thinking goals left implicitmuch reduced effect

The last row is the one that demolishes most corporate training: if critical thinking is an implicit objective (“you pick it up on the job”), the measured effect is small. It has to be stated, practised and corrected.

AI moved the problem, it didn’t solve it

There’s a reason this article also sits in the AI category.

In 2025 a Microsoft research group led by Hao-Ping Lee published a survey at the CHI conference of 319 knowledge workers using generative AI, collecting 936 concrete usage examples. The reported result: the more participants declared confidence in the tool’s ability, the less critical effort they declared applying; the more confidence they declared in their own skills, the more they applied. It’s a self-report survey, so it doesn’t establish causation, but the direction is consistent with what is observed elsewhere.

Fabrizio Dell’Acqua at Harvard Business School had described a similar phenomenon in a 2023 experiment with recruiters: when the algorithm was very good, humans stopped checking and overall results got worse. He calls it falling asleep at the wheel.

The practical version: the cost of producing a plausible text has collapsed, the cost of verifying it hasn’t. That’s why how you use these tools matters more than which one you use. I’ve written about it in how to use ChatGPT at work and how to write effective prompts.

How to actually train it

  1. Build domain knowledge first. There’s no shortcut. Critical thinking applied to a field you don’t know is indistinguishable from generic suspicion.
  2. Look for the strongest argument against your position, not the weakest. It’s the operational antidote to myside bias, and it’s an exercise you can do in writing in ten minutes.
  3. Ask yourself what observation would change your mind. If the answer is “none”, you’re not reasoning about a fact, you’re defending an identity.
  4. Slow down on answers that arrive immediately. The CRT doesn’t measure intelligence, it measures willingness to distrust your first impulse.
  5. Argue in a structured way with someone who knows the field. That’s the combination producing the largest effect in the meta-analysis. Alone you improve much less.
  6. With AI, always verify at least one verifiable thing. A number, a citation, a name. If you can’t find a single thing to check, the text probably says nothing.

My take: in job ads it has become an empty word

I think critical thinking has become an umbrella term that in job ads means “we’d like someone who doesn’t cause disasters”, and that this makes it practically unassessable in an interview. The questions often used to measure it, puzzles and abstract cases, mostly measure familiarity with that kind of puzzle.

I also have the impression, and this really is only an impression, that the most visible effect of generative AI in the coming years won’t be on production but on verification: when everything is plausible, the competitive advantage shifts to whoever can tell the difference. I have no data to support this, and it counts as a personal forecast, not a conclusion from the literature.

Finally: I find Kahan’s result the most important in this whole literature and the least cited. The idea that the solution to misinformation is “more skills” is comforting and, at least in that study, false.

Frequently asked questions

So critical thinking courses are useless? No, but effectiveness depends on format. The ones with structured dialogue, real problems and supervision work; the theoretical ones disconnected from work, far less.

If intelligence doesn’t help, what does? In Stanovich’s literature the variable that matters most isn’t cognitive ability but the disposition to think open-mindedly: the willingness to seek out contrary information. It’s a disposition, not a talent.

How do I notice I’m in myside bias? A reliable signal is the double standard: if you accept a statistic that supports you instantly and go check the methodology of one that doesn’t, you’re in it.

Can critical thinking be measured? Instruments exist (Watson-Glaser, California Critical Thinking Skills Test) and they’re the ones used in many of the meta-analysis studies. They correlate with academic outcomes, less clearly with job performance.

Is it worth studying logical fallacies? As vocabulary yes, as method not much. Recognising a fallacy’s name isn’t the hard part; the hard part is noticing that you’re committing it.

Sources

This article was written following the process described in how we write. If you find a figure reported incorrectly, write to me: I’ll correct it and flag it at the bottom of the article.

Founder of UpSkillLab, with a Master's in Digital Marketing from the University of Florence. Every article starts from public sources cited in the text and is drafted with the help of AI tools, under my review and editorial responsibility: no purchased content, nothing published without human checking. Read the editorial method or report an error.


Share this post:

See all articles in Skills & Competencies

Previous Post
English for Work: 2,000 Words Cover 95% of What You Hear
Next Post
Personal Branding: A Signal Anyone Can Copy Isn't a Signal