The critical thinking toolkit
This is a set of tools you can use to test any claim, whether it comes from economics, science, politics, advertising, social media, or even an argument with your own family.
Seven tools, one checklist
Each tool helps you test the claim in front of you from a different angle and gets you closer to deciding whether the claim is trustworthy.
Allow yourself these three things
Thinking critically is not about always having the right answer. These three are part of the process, not signs of failure.
Not knowing yet
"I don't know — what evidence would help me decide?" is a stronger answer than a confident guess. It is what good researchers say constantly.
Holding a view without complete confidence
You don't need to be certain before you act. Most decisions are made under uncertainty, by people who know they might be wrong.
Changing your mind
It is the only way you ever get closer to being right.
Break the claim into five parts
When someone makes a strong claim, take it apart before you agree or disagree with it.
- Claim: What, exactly, is the person saying?
- Evidence: What facts or data are offered in support?
- Reasoning: How is the evidence supposed to prove the claim? This is where most weak arguments break.
- Assumptions: What has to be true for the reasoning to work? Assumptions are rarely stated out loud — you have to find them yourself. It is the step most people skip.
- Counterargument: What would a thoughtful person who disagrees say?
Here is one claim taken apart. The middle column is what the person making the claim says (or quietly relies on). The right-hand column is the question you ask back.
| Part | What the claim-maker says | What you ask |
|---|---|---|
| Claim | "AI will cause mass unemployment." | How many people counts as "mass"? By when? |
| Evidence | "AI can already do many of the tasks in office jobs." | Is employment actually falling in those jobs yet? |
| Reasoning | "If a machine can do the task, the worker isn't needed." | Is doing some tasks the same as doing the whole job? |
| Assumption | Not said out loud: new kinds of work won't appear fast enough to employ the people who lose theirs. | Has that been true in past waves of new technology? |
| Counterargument | — | ATMs took over bank tellers' main task, yet the number of US tellers kept growing for years afterwards, because cheaper branches meant banks opened more of them. How would the claim-maker answer that? |
Try it: "Homework is useless — my cousin never did any and she got into a great university." Find the evidence and the hidden assumption.
Evidence: one person's story. Assumption: that what happened to one person is what usually happens — and that she succeeded because she skipped homework, not despite it. A good question back: what happened to the students who skipped homework and didn't get in? You never hear their stories.
The habit to build
Surface the assumption underneath a claim — then test whether, and when, that assumption actually holds.
Ask "compared with what?"
A number on its own means almost nothing.
Suppose someone announces that a new technology will create one million jobs. That sounds excellent. But:
- How many jobs would have been created anyway?If 500,000 would have appeared regardless, the real contribution is half of what was claimed.
- How many jobs did it destroy at the same time?Create a million, destroy 800,000, and the net effect is 200,000 — a fifth of the headline.
The same question punctures claims you'll see every week:
A tuition centre ad: "Our students improved by two grades."
Compared with students who didn't go? Many students improve over a year anyway, as they revise and the syllabus clicks.
"This medicine helped 60% of patients."
Compared with a sugar pill? For some conditions, a large share of patients feel better on a placebo too. The medicine only deserves credit for the difference.
Try it: "9 out of 10 dentists recommend this toothpaste." Compared with what?
Were the dentists asked to pick one brand, or allowed to name several? If each dentist could recommend many brands, almost every toothpaste could make the same claim. The number sounds like a contest result, but there may have been no contest.
The habit to build
Every impressive number needs a comparison attached before it means anything at all.
Ask "who gains, and who pays?"
This tool is about groups. Almost no change is good or bad for everyone. It is good for some people and costly for others, and the interesting question is for whom.
When you hear that something will make society richer, safer or more efficient, ask where the benefit actually lands. It could go to any of these — and they are very different outcomes:
- Consumers, through lower prices
- Workers, through higher wages
- Companies, through higher profits
- Shareholders, through rising share prices
- Governments, through more tax revenue
A country can become measurably more productive while most workers see almost nothing extra in their pay. Both statements can be true at once — and if you only track the first, you will be baffled by how angry people are about the second.
Try it: Your school bans phones for the whole school day. Who gains, and who pays?
Possible gainers: teachers (fewer distractions), students who wanted an excuse to switch off, parents worried about screen time. Possible payers: students who use their phone to arrange getting home, students who relied on it to keep up with friends, and the staff who now have to enforce the rule. Your list may differ — the point is that "the school benefits" hides several very different groups.
The habit to build
"Society benefits" is never a complete answer. Ask which part of society.
Ask "over what period?"
Tool 3 asked who gains. This tool is about time: the same change can look very different next month and in ten years.
Resist the pull towards a single verdict. A new technology can be good for productivity, good for consumers, good for some companies, bad for other companies, bad for particular workers, and unsettling for society — all at the same time. None of those cancel each other out. They are just different questions.
And time matters as much as perspective. Short-term and long-term effects can point in opposite directions:
Some jobs disappear. Workers need retraining. Particular industries struggle. Some companies profit handsomely.
New industries develop. New occupations appear. Productivity rises. Prices may fall, and living standards may rise.
The trap hiding in those two boxes
Both boxes can be accurate — but long-term benefits do not automatically justify short-term suffering. Watch for the moment an argument quietly shifts from what will happen ("things will be better in twenty years") to what we should do ("so the people hurt now should just accept it"). The first is a prediction you can check with evidence. The second is a moral judgement, and it needs different kinds of reasons.
Try it: "Cramming the night before a test works — I always pass." Over what period?
Short term, cramming can get you through tomorrow's test. Long term, research on memory consistently finds that practice spread out over days is remembered far better than the same practice squeezed into one night. So "it works" is true for one test and misleading for the final exam that covers everything.
The habit to build
Replace "is this good or bad?" with "good for whom, in what way, over what period?"
Separate the individual from the group
Tool 3 looked at groups and Tool 4 at time. This tool is about scale: one person's experience versus the effect on everyone. Consider a taxi driver when self-driving cars arrive.
Fewer jobs, possibly none. Very bad.
Cheaper transport, more people travelling, new services built on top, possibly new jobs elsewhere. Plausibly good.
Both are true. The error is using one to dismiss the other — telling the driver the economy is fine, or insisting that because his situation is bad the whole change must be.
Try it: a thought experiment
You run a company with 1,000 employees. A new system lets you do the same work with 700.
What do you do?
- Let 300 people go
- Keep everyone and produce far more
- Keep everyone, retrain them for new tasks
- Cut working hours, hold pay steady
- Cut prices sharply and grow demand instead
Decide. Then answer again as one of the 300. Then as a shareholder. Then as a customer. Then as the government. Then as a student about to enter that industry.
Did your priorities change? That is the whole lesson: different people can look at exactly the same event and have completely different interests.
The habit to build
Before judging a decision, name whose chair you are sitting in.
Apply the "this time is different" test
Whenever someone says a change is unprecedented, be careful — in both directions.
People have predicted the end of work repeatedly. In 1900, about 40% of American workers were in farming; today it is under 2%, and mass permanent unemployment never arrived. Whole occupations vanished, and whole new ones appeared that nobody could have named in advance. That history should make you sceptical of doom.
But "it worked out before" is not proof it will work out again. That is a pattern, not a law.
So ask precisely
What is genuinely different this time, and is that difference big enough to break the old pattern?
Then the follow-up: even if the pattern holds and new jobs eventually appear, do they appear fast enough, in the right places, needing skills the displaced people can realistically acquire? The gap between "eventually fine" and "fine for the people living through it" is where most real policy arguments actually live — and you can watch both halves of that question being put to a real argument in the first case study.
Try it: "Adults panicked about comics, then TV, then video games. Social media is just the latest panic." Apply the test.
The history is real, so it's a fair challenge. Now name what might be genuinely different: social media is with you all day, not just when you sit down to watch; it involves comparing yourself with people you actually know; and it is designed to keep you scrolling. Then ask whether any of those differences is big enough to change the outcome. That is exactly the argument in the second case study.
The habit to build
Name the specific difference, then ask whether it is big enough to change the answer.
Ask "what would change my mind?"
This is the tool that separates thinking from cheerleading.
For any position you hold, you should be able to describe the evidence that would make you abandon it. If you can't, you don't hold a position — you hold a loyalty.
The single most useful move in this whole guide
When two explanations both fit the facts you have, don't pile up more evidence for your favourite. Hunt for the place where the two explanations predict different things, and look there.
Say exam scores dropped the year your school changed its timetable. One explanation: the timetable caused it. Another: the exam was simply harder that year. More evidence that scores fell won't help — both explanations predict that. But they disagree about other schools. If the timetable is to blame, only your school should have dropped. If the exam was harder, every school should have dropped. So check the other schools.
That one move settles more arguments than any amount of debating — you can watch a student go looking for the rival explanation in the first case study.
Try it: "We lost because the referee was biased against us." What would change your mind?
Some honest answers: replays showing the key calls were correct; the same referee giving a similar number of fouls to both teams across many matches; your team losing by the same margin in games with other referees. If you can't name anything that would count, it's a loyalty, not a position — which is fine for a fan, but worth noticing.
The habit to build
Before defending a view, state what would defeat it.
The checklist for evaluating a claim
Eight questions that pack the seven tools into one list. Run them on anything — a headline, an advertisement, a classmate, a politician, this page.
- What exactly is the claim? State it in one sentence, in your own words. (Tool 1)
- What evidence supports it? Facts and data — not confidence, not volume. (Tool 1)
- What assumptions is it resting on? The unstated things that must be true for the reasoning to work. (Tool 1)
- Compared with what? Every number needs a baseline. (Tool 2)
- Who gains, and who pays? Name the groups, not "society" — and don't let one person's story stand in for everyone's. (Tools 3 and 5)
- What would the strongest opponent say? The strongest version, not the easiest one to beat. (Tool 1)
- Does the answer change over the short term versus the long term? They often point in opposite directions. (Tool 4)
- What evidence would change my mind? If nothing would, it isn't a position. (Tool 7)
Tool 6 isn't on the list because it only applies to one kind of claim. Reach for it whenever someone says a change is unprecedented — or that it's just like last time.
Practise it this week
Pick any confident claim you meet — an advertisement, a headline, a TikTok, a classmate, a politician. Start with just three questions: What exactly is the claim? Compared with what? What would change my mind? Write the answers down — it is harder than doing it in your head, which is precisely why it's worth doing. Once three feel easy, add the rest of the checklist.
The four blanks
Complete it honestly — especially the last one.
"I used to think ______. Now I think ______, because ______. I'm still unsure about ______."
If that last blank is empty, you probably haven't thought hard enough yet.
See the tools used on real arguments
The tools are easy to nod along to and much harder to actually use. Each case study takes one live, unsettled disagreement and walks the toolkit through it. And when you've seen them work, put the toolkit itself on trial.