Practical Rules for Critical Thinking

A practical set of rules for critical thinking across foundation, analysis, and engagement, written for non-academics.

I’m seeing a lot of references to “critical thinking” in my social media feeds, but often those most excited about critical thought also hold ideas that are strong evidence of its absence. “Think critically! Examine everything! The world is flat!” Those people define “critical thinking” differently than I do.

I’ve organized my working rules into three domains:

1. Foundation Rules: The attitudes and intellectual humility required for critical thinking.
2. Analytical Rules: Methods for evaluating information, weighing evidence, and reaching conclusions.
3. Engagement Rules: Testing and refining our thinking through reasoned discussion with others.

I’ve examined ICAT, Bloom, Paul-Elder, and CARS only superficially (see Analytical Rule One). I wanted a set of guidelines non-academics could use.

Foundation Rules

Mindset and fundamental approach. The attitudes and intellectual humility required for critical thinking.

Foundation Rule One: Assume you could be wrong.

Critical thinking begins with skepticism: question and doubt by default. First doubt your own position. Your current thinking may be flawed, and you cannot examine it critically if you refuse that possibility. Socrates expressed the attitude: “I neither know nor think I know.”

Foundation Rule Two: Identify the circumstances under which you’d change your mind.

After admitting you could be wrong, identify evidence that would change your mind. The threshold may be high. As Carl Sagan said, “Extraordinary claims require extraordinary evidence.” But there must be a threshold.

Foundation Rule Three: There is nothing wrong with changing your mind.

Changing a belief need not change your identity. Treat a challenge to an idea as a challenge to that idea, and accept being wrong initially if you can correct it.

Foundation Rule Four: It’s always better to be really right than mostly right.

A useful model can still be inaccurate. The Ptolemaic geocentric model served astrologers and eclipse predictions, but further investigation produced knowledge that took us to the moon and will likely take us to Mars. Stopping at today’s useful approximation limits the problems we can solve later.

Precise measurements of gravity and the Earth’s shape provide two further examples.

How can you find out how many geophysicists are in the room? Announce that the acceleration of gravity on Earth is 9.8 m/s^2. “At the equator, at sea level, if there’s no large ore bodies nearby, sure. But everywhere else, uh, nope.” Geophysicists use the precise measurement of gravitational variations to find ore bodies. That you’re reading this on a computer is thanks, in part, to the pursuit of “really right” measures of gravity.

And how can you find out how many geographers are in the room? Announce that the earth is an oblate spheroid. “If you ignore the oceans, flatten all mountains, and allow a margin of error wide enough to fit the Grand Canyon, sure. And even then it’s more of a lumpy pear shape with dynamic tidal deformations.” If the only use for measuring the Earth is to find your way from Spain to Cuba, mostly right is good enough. But not good enough to get Google Maps to show you the way home.

Foundation Rule Five: Reserve judgement.

Examine evidence before reaching a conclusion. Stereotypes and claims such as “all X are Y” can prevent analysis, often for emotional reasons. Starting with “everyone is corrupt” may sound skeptical, but it settles the conclusion in advance. Conspiracy theories depend on this sort of cynical prejudgment.

Foundation Rule Six: All generalizations are wrong, including this one.

Be as specific with your argumentation as possible and as general as necessary. Be careful with broad generalizations because they tend to include unintended exceptions. Most human characteristics, behaviors, and situations exist on a spectrum rather than in distinct categories, so sweeping statements usually capture cases they shouldn’t or miss cases they should include.

Foundation Rule Seven: Read more that challenges you than confirms you, and don’t believe everything you read.

Seek well-reasoned challenges to your views and question confirming sources as well. We tend to prefer information that validates existing beliefs, so spend more time understanding opposing arguments than reinforcing your position. This requires the courage to examine possible errors and the skepticism to question familiar sources. Otherwise, the remaining rules can become ways to defend existing beliefs.

Analytical Rules

The methods and principles we use to evaluate information, weigh evidence, and reach conclusions.

Analytical Rule One: Remember that for effectively every human endeavour, there are probably at least 100 people with PhDs in it.

Room colour and happiness, or baking sourdough at high altitude, may each occupy a dozen PhD students. Nearly every subject has people who have spent decades studying it. Unless you are the leading expert, someone probably knows far more. Before examining a topic critically, establish a basic understanding and identify the relevant academic disciplines.

Analytical Rule Two: If your opinion differs from those who know the subject matter better than you do, you’re almost certainly wrong.

This is an idea called epistemic humility. For example, if there are 1000 researchers investigating how wood decomposes across variations in humidity, temperature, and local micro-organisms and they all agree but you don’t, it’s almost certain you are wrong. If there are two opinions on the matter and the groups are evenly split, it’s reasonable for you to disagree with one side but not both. If there are 1000 opinions on the matter but 100,000 experts agree on one and 999 others each have their own opinion, it’s not reasonable to disagree with the effective consensus opinion. Assuming independent and distinct investigation and judgement, the likelihood of error in a consensus decreases exponentially as the number of experts supporting it increases. This principle applies most strongly to established fields with mature methods and evidence. In emerging fields, consensus may be more fluid, and further investigation is often needed. But it’s essential to seek the predominant view among experts, rather than selectively finding experts who happen to align with your initial opinion.

This is where “doing your own research” is a bit misleading. No matter how much amateur research you do, you’re not going to surpass the collective knowledge of the experts in the field. The “research” you should be doing is finding out what the smartest, best-informed in the field have to say on the matter. If a plurality of them agree on one position, it’s irrational for you to disagree.

An undeniably brilliant acquaintance studied at MIT, earned a Harvard MBA, and builds proprietary financial models that make him and his firm a lot of money. He uses enterprise computing for large simulations and understands financial modeling’s capabilities and limitations. He mistakenly assumes that this expertise applies to another scientific field, arguing that its models cannot justify the certainty scientists claim. Roughly 20,000 scientists support the claim he disputes; roughly 600 dissent, almost all with different positions. That distribution gives the consensus enormous statistical weight. The question here is whether his doubt is reasonable, irrespective of whether he is right.

Even with conservative estimates, his judgment would need the statistical weight of more than 2,000 experts to give him even a 10% chance of being right. He could be correct, but the odds are about as likely as winning the lottery three times in a row. It is not rational to believe he is right and all the experts are wrong. Even if he ranks in the top 0.1% by intelligence, making him the smartest in a thousand, he would not statistically rank in the top ten among 20,000 scientists. It is unreasonable to assume that those with greater intelligence, experience, and subject knowledge are collectively wrong and he is right.

Analytical Rule Three: Learn how reasoning works and fails; use it to understand rather than score points.

Learn how deduction applies general rules to specific cases and induction derives general rules from cases. Watch for confusion between correlation and causation, attacks on people, and conclusions based on too little evidence. Recognize confirmation bias and the availability heuristic, which gives easily remembered information too much weight. A flawed argument can still reach a correct conclusion. Use this knowledge to detect misleading arguments and improve your own reasoning, while remembering Engagement Rule Three: people with useful ideas may lack rhetorical skill, and skilled rhetoricians may intend to mislead.

Analytical Rule Four: Sometimes there are not enough facts or too many opinions to reasonably find a single good answer.

Limited information or opportunities for experimentation can leave a field with competing theories and opinions. A thousand experts defending a thousand positions probably indicates few established facts. Then examine the methods, measures, researchers, and evidence more closely. Give tested findings more weight than untested claims, and look for consilience: agreement among independent, diverse methods and fields. I have more detailed rules for this work, but it is easy to undertake it unnecessarily. Where most experts agree, they have already examined these questions, and an independent amateur review is very unlikely to add value.

Elon Musk has become the world’s richest man partly by challenging experts. He funds scientists to design and run experiments that produce new evidence. That differs from an amateur reinterpreting existing findings. His approach also illustrates the value of greater accuracy: he persists beyond useful approximations until new findings change what can be done. I am not a Musk fan, but acknowledge his success.

Analytical Rule Five: Even a consensus could be wrong. But “wrong” in this case usually means “not really right” and not “totally wrong.”

As the saying goes, “science advances one funeral at a time.” Scientists are people too so they become attached to their own work to the point of disregarding evidence that invalidates or challenges it. However, Newtonian physics is not “wrong” until you increase the scope of observations. Only when you need incredible precision or you leave the planet do you need relativity. Science, for the most part, advances by gradually uncovering the truth. The popular analogy is of a sculptor chiseling to reveal the truth within the block of stone. Each chunk of stone removed represents additional insight into what lies beneath. Very rarely does the insight represent a huge change and in almost all cases is a refinement or enhancement of previous thought.

Analytical Rule Six: Test what you can, but think critically about your testing. And remember not everything can or should be tested.

Testing connects theory to observation and is how Musk makes his billions. Scientists still need to decide what to test, define success and failure, select representative conditions, and examine variables and assumptions. Someone saying “I don’t think, I just test” is using empirical testing as a framework for critical analysis, whether they recognize it or not. Tests require thought about their design and limits. Many things cannot be tested practically or ethically. We cannot field-test macroeconomic theories with control variables because there is only one global economy. We cannot test most social sciences; one of many reasons is that we cannot ethically raise feral children as experimental controls. Identifying the limits of empiricism itself requires critical thinking. Testing is useful but necessarily cannot be the only approach.

Engagement Rules

Testing and refining our thinking through reasoned discussion with others.

Engagement Rule One: Test your opinions against the best examples of their counter-arguments.

This practice is called “steel-manning”: seek the best-informed criticism of your position. If it changes your mind, apply Foundation Rule Three.

Engagement Rule Two: If you’re debating, argue about ideas not words. Unless you’re arguing about words, that is.

Make sure that when you are discussing with someone that you are using the same language. Imprecise terms and definitions are a time waster. And if someone refuses to define terms or changes their definitions mid-stream for whatever reason, find someone else to discuss with. They’re not arguing in good faith.

Engagement Rule Three: Be generous. Not everyone with a valuable opinion is a skilled communicator. Apply the most charitable interpretation you can to an argument rather than the most restrictive.

Misinterpreting an argument because someone misspoke or misused a technical term is disingenuous. Useful ideas need not be eloquently expressed. Apply the same standard to yourself: state your argument carefully enough to reduce the need for charitable interpretation. Pedantry can be useful, but not when it prevents understanding.

Engagement Rule Four: Being right is the objective. Winning is ending up right, not proving the other guy wrong.

Use discussion to improve your understanding rather than to win a debate. Critical thinking requires both skill and a commitment to pursuing knowledge with integrity.

Engagement Rule Five: Assume good faith but pick your battles.

Consider ideas fairly regardless of their source. Begin by assuming inquiry or observation rather than propaganda, foolishness, or malice. Even a rejected idea may contain something useful. If examination reveals bad faith, you can stop engaging. You may leave any conversation at any time.

Engagement Rule Six: Regularly assess how well you’re doing at critical thinking.

Evaluate your thinking against these criteria:

• Can I articulate opposing views fairly?
• Have I identified my assumptions?
• What thinking habits need improvement?
• Where did I show bias?
• What should I do differently?

I wrote these rules mainly for myself and to have a reference I could share. You can share or reuse them without a license or attribution.