Philosophy › Section B · Critical thinking, reading, and writing › Topic B.2

Overcoming cognitive biases

The shortcuts that make thinking fast also make it lopsided in predictable ways. Knowing the common biases by name, and knowing which defences actually work, is the first practical step towards reasoning like a philosopher.

From shortcut to bias

The previous topic, The brain as an inference machine, showed that the mind relies on fast heuristics that usually serve it well. A cognitive bias is what happens when such a shortcut fails in a systematic way: a predictable tilt away from what the evidence and good reasoning support. Biases are close cousins of the errors in the topic Informal fallacies, but not the same thing. A fallacy is a defect in an argument, something you can point to on the page. A bias is a tendency in the thinker, shaping which arguments you notice and how hard you test them.

Metacognition: thinking about thinking

The main defence is critical reflection, often called metacognition: monitoring and steering your own thinking. Ordinary cognition is first-order, thinking about the world; metacognition is higher-order, thinking about how you are thinking. In practice it means checking your work, planning before you start, asking how well you really understand a passage, and noticing when you have leapt to a conclusion. Under stress, time pressure or constant interruption, people fall back on quick answers, so unhurried, undistracted time is part of the method.

Confirmation bias

Confirmation bias is the tendency to seek, interpret, favour and remember information that supports what you already believe. As a default it is often sensible: hearing rain, you look out of the window for rain, not for proof that it is sunny. It becomes a bias when it decides the question for you. The psychology topic Problem solving describes Peter Wason's 1960 task, in which most people asked to discover a rule fitting the numbers 2, 4, 6 tested only sequences that fitted their guess, and so failed to find that the rule was simply any rising sequence. The cure Wason's task suggests is falsification: look for the case that would show you are wrong.

Confirmation bias online

It is often said that recommendation algorithms trap people in bubbles that harden their views. The evidence is more mixed. In a large experiment reported in Science in 2023, Andrew Guess and colleagues switched consenting Facebook and Instagram users to a simple chronological feed for about three months during the 2020 US election campaign; their measured political polarisation did not change significantly. One study of one platform family and period does not show feeds are harmless, but it suggests that algorithms are not the whole story: much selective exposure also comes from our own choices of what to follow, click and share, which is where confirmation bias lives.

Anchoring and availability

Anchoring is the pull of a first number on later estimates, even when the number is plainly arbitrary. In Amos Tversky and Daniel Kahneman's 1974 report, a spin of a wheel of fortune shifted people's estimates of the percentage of UN members that were African countries. Anchoring has replicated well in large multi-laboratory projects. The availability heuristic judges how common something is by how easily examples come to mind, so dramatic or heavily reported events seem more frequent than they are. In philosophy, the first view you meet on a question can quietly anchor what you think afterwards.

Tribalism and motivated reasoning

We readily side with groups that share our values. In the early 1970s Henri Tajfel showed that even groups formed on trivial grounds, such as a stated preference between two painters, were enough for people to favour their own group when sharing out rewards. Tribalism makes it hard to judge fairly an argument that comes from the other side. A close relative is motivated reasoning: reaching the conclusion you want and then gathering reasons for it. The bandwagon fallacy, believing something because many others do, trades on the same pull: popularity may be a clue, but never makes a claim true.

Sunk costs and the gambler's fallacy

Sunk costs are time, money or effort already spent that cannot be recovered. The sunk cost fallacy treats them as a reason to keep going: staying in a failing venture because of what you have put in, rather than because of what it will return. The gambler's fallacy is the belief that a chance event is due because it has not happened lately. Both look backwards when the decision is about the future. Philosophers note that the past is not always irrelevant: commitments and loyalty matter, and Seth Lazar (2018) has asked whether harm already done should count when judging whether a war remains justified.

Worked example: a run of heads

A fair coin lands heads five times in a row. Maya says tails is now "due". 1. Each toss of a fair coin is independent: the coin has no memory, so the chance of heads on the next toss is still \(\frac{1}{2}\). 2. The feeling comes from the whole run: before any tosses, six heads in a row had chance \((\frac{1}{2})^6 = \frac{1}{64}\). 3. But five heads have happened; only one toss is uncertain. 4. Maya confuses the improbability of the whole sequence, judged in advance, with the probability of the next step. 5. Verdict: gambler's fallacy. The same habit of reasoning from the past, not the future, drives the sunk cost fallacy.

Worked example: naming the bias

Classify each case. 1. Leo has spent two years on a degree he now dislikes and says he must finish "or the two years were wasted": sunk cost fallacy, since the question is which future is better. 2. After reading one striking news story about a shark attack, Priya decides swimming in the sea is too dangerous: availability, as one vivid case stands in for the actual rate. 3. A seminar member rejects an argument as soon as he learns it was made by a rival political group: tribalism. 4. Sam reads only the reviews that praise the phone he has already chosen: confirmation bias. 5. Check: ask what each decision should depend on, and what it did depend on.

Do the defences work?

Knowing the names of biases is not enough. Emily Pronin and colleagues described a bias blind spot: people readily see bias in others while rating themselves as less biased. What helps more is changing the procedure. In a 1984 study, Charles Lord, Mark Lepper and Elizabeth Preston found that asking people to consider the opposite, imagining how they would judge the same evidence had it pointed the other way, reduced biased assimilation of evidence. Slowing down, writing your reasoning out, setting criteria in advance and inviting criticism also help. Research on how long debiasing lasts is still developing, so treat these as aids, not guarantees.

Common mistakes to avoid

First, thinking biases afflict only careless people: they are built into normal cognition. Second, confusing a bias, a tendency in the thinker, with a fallacy, a defect in an argument. Third, assuming that learning the list protects you. Fourth, treating every glance at the past as a sunk cost error: commitments can be genuine reasons. Fifth, believing a chance event becomes due after a run of the opposite outcome. Sixth, repeating confident slogans about algorithms without looking at the evidence.

Key takeaways

A cognitive bias is a systematic tilt from a normally useful shortcut; a fallacy is a defect in an argument. Metacognition is the main defence and needs time and calm. Confirmation bias favours existing beliefs; seek disconfirming cases. Anchoring lets a first number pull estimates; availability lets vivid cases stand in for rates. Tribalism and motivated reasoning bend judgement towards our group and our wishes. The sunk cost and gambler's fallacies look to the past when deciding about the future. Procedures such as considering the opposite help more than knowing the names.

Flashcards

Flashcard

Cognitive bias

A systematic tilt away from what evidence and good reasoning support

A useful shortcut failing predictably.

Test your understanding

Question 1 of 5

Ana paid $80 for a non-refundable concert ticket. On the night she feels ill and would rather stay home, but goes so the money is not "wasted". Which error is this?

Question 2 of 5

A fair coin has just landed heads five times in a row. What is the probability that the next toss lands heads?

Question 3 of 5

Which practice most directly targets confirmation bias?

Question 4 of 5

What did Henri Tajfel's experiments of the early 1970s suggest?

Question 5 of 5

How does a cognitive bias differ from an informal fallacy?

← Previous topicNext topic →