The Biased Brain — Critical Thinking 5 (C1)

CRITICAL THINKING · LESSON 5

The Biased Brain

C1 (B2–C2) · 90 minutes

Your estimate

Please don't read beyond this slide before class. The first three slides are experiments, and you are the subject.

Task Your teacher will send you a number privately. Don't share it. Write your answers down before anyone speaks.

1. How many living languages are spoken in the world today? Is it more or less than the number your teacher sent you?

2. Write your best estimate of the exact number. Send it to your teacher in a private message.

Meet Linda

Task Read the description and write A or B. Don't discuss it yet.

"Linda is 31 years old, single, outspoken, and very bright. She majored in philosophy. As a student, she was deeply concerned with issues of discrimination and social justice, and also participated in anti-nuclear demonstrations." (Tversky and Kahneman, 1983)

Which is more probable?

A

Linda is a bank teller.

B

Linda is a bank teller and is active in the feminist movement.

What comes to mind?

Task Answer quickly. Don't look anything up.

1. Take English words of three letters or more. Is the letter K more likely to be the first letter of a word, or the third? Write first or third, then how many times more likely (for example, 2 to 1).

2. Worldwide, in a typical recent year, how many people die by drowning for every one person killed by terrorism? Write a number.

The reveal

16,000

1. Did a number you knew was arbitrary still move your estimate? Why?

2. If you chose B for Linda, what were you really answering: how likely or how typical?

3. What made the K question or the drowning question feel easy to answer?

14,000

12,000

10,000

8,000

Ethnologue:

just over 7,000

6,000

4,000

2,000

0

Group given 1,500

Group given 15,000

READING

Two systems

Before you read, ten seconds only: 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?

In Thinking, Fast and Slow (2011), the psychologist Daniel Kahneman (1934–2024), winner of the 2002 Nobel Prize in economics, describes the mind as a partnership of two modes of thought. System 1 is fast, automatic and effortless. It recognises faces, completes "bread and…", reads hostility in a voice, and produces an answer before you ask for one. System 2 is slow, deliberate and effortful. It multiplies 17 by 24, checks an

argument, and follows a rule against your instincts.

Most of the time System 1 runs the show and System 2 accepts its suggestions. That is usually sensible, since System 1 is quick and often right. The trouble is that System 1 answers whatever question it can answer, and System 2 is lazy: it rarely checks. The labels themselves came from the psychologists Keith Stanovich and Richard West (2000). Kahneman stressed that the two systems are useful fictions, a way of telling the story, not two parts of the brain.

READING

Three shortcuts

In 1974 Amos Tversky and Daniel Kahneman published "Judgment under Uncertainty: Heuristics and Biases" in Science. Their claim was that people judge uncertain questions with a few mental shortcuts, or heuristics, which usually work but produce biases: errors that are systematic, not random.

Anchoring is the pull of a starting number. In one study a wheel marked 0 to 100 was fixed to stop at 10 or at 65. Participants then estimated the percentage of African countries in the United Nations. Those who saw 10 gave a median answer of 25; those who saw 65 gave 45. People adjust away from an anchor, but not far enough.

Availability is judging how common

something is by how easily examples come to mind. In 1973, 105 of 152 participants judged that most of five letters, including K, appeared more often first in English words than third. All five appear more often third. Words are easier to retrieve by their first letter, and vivid events (a plane crash, an attack) are easier to recall than common, quiet ones.

Representativeness is judging probability by resemblance to a stereotype. In 1983 the pair gave the Linda problem to 142 students at the University of British Columbia: 85% judged "bank teller and feminist" more probable than "bank teller". Yet a conjunction can never be more probable than one of its parts. They called this the conjunction fallacy.

Three shortcuts

1. Where in your own country's news does availability shape what people fear?

2. Is choosing B for Linda irrational, or a sensible reading of an odd question?

READING

Seeing what we want to see

Confirmation bias is the tendency to look for, notice and remember evidence that fits what we already believe. In Peter Wason's "2-4-6" task (1960), people had to discover a rule that the triple 2, 4, 6 obeyed. Most tested only triples that fitted their own guess, such as 8, 10, 12, and rarely tried one that could prove them wrong. The rule was simply "any three increasing numbers".

Motivated reasoning adds a motive: we reason towards the conclusion we want.

The social psychologist Ziva Kunda (1990) argued that we are not free to believe anything; we must still build a case that looks reasonable. But wanting a conclusion changes which evidence we search for and how hard we test it. In a study by Lord, Ross and Lepper (1979), supporters and opponents of the death penalty read the same mixed evidence. Each side rated the study that agreed with it as more convincing, and both sides left more certain than before.

READING

The Lisbon hire

A fictional case. Marta Oliveira founded Tejo Freight, a 60-person logistics software company in Lisbon. She needs a head of engineering.

1. At a conference in Berlin, Marta hears Arjun Mehta give a dazzling talk. He is confident, quick and fresh from a famous tech firm: exactly what she imagines a head of engineering looks like. Ten minutes into their coffee, she has decided.

2. Her last head of engineering, a quiet and careful woman, left after a disastrous launch. The story is retold at every board meeting. Marta quietly drops two strong internal candidates because "they're too quiet".

READING · CONTINUED

The Lisbon hire

3. Arjun asks for 210,000 euros, well above the company's 140,000-euro band. Marta bargains hard and agrees 185,000 euros. She feels she has won.

4. In the interview she asks about his successes. A board member, Kofi Mensah, mentions a product Arjun led that failed. Marta decides Kofi is jealous.

5. She calls the two referees Arjun chose. One says, "Brilliant, though

hard to work with." Marta's notes say "Brilliant".

6. After three months two senior engineers resign, both naming Arjun. Marta has already called the hire "my best decision of the year" in a press interview. She concludes they couldn't handle high standards.

7. After nine months the main project is half a year late, and Arjun leaves.

Post-mortem

Task Copy the table. For each step from 1 to 6, name the bias (or biases) and the System 1 shortcut behind it. Then propose one safeguard per step: a rule, habit or procedure Tejo Freight could adopt before the next hire. Finally, choose the single safeguard that would have saved the most money.

Step

Bias (or biases)

One safeguard

1

2

3

4

5

6

READING

Handle with care

This field has its own warning attached. Psychology's replication crisis, which came to a head in the 2010s, found that many published results did not appear again when other laboratories repeated the studies. Thinking, Fast and Slow drew on several of them. Ego depletion, the idea that self-control draws on a limited supply that runs down with use, was tested by 23 laboratories with 2,141 participants in 2016; the effect was close to zero, though its defenders dispute the test used. Behavioural priming, the claim that reading words about old age makes people walk more slowly, failed to

replicate in 2012. Kahneman wrote to priming researchers that year warning of a "train wreck looming", and later admitted that he had "placed too much faith in underpowered studies". By contrast, anchoring has replicated strongly in large multi-laboratory tests.

Nor are heuristics simply flaws. Gerd Gigerenzer argues that simple rules are often well suited to their environment. Asked which city is larger, San Diego or San Antonio, 62% of American students chose San Diego, the right answer; every German student did, often simply because it was the only one they had heard of.

Bias in the wild

Task Rate each statement privately: 1 strongly agree · 2 mostly agree · 3 on the fence · 4 mostly disagree · 5 strongly disagree. Then compare. Spend most of your time on the two statements your group disagrees about most.

1

strongly agree

2

mostly agree

3

on the fence

4

mostly disagree

5

strongly disagree

1. Knowing about a bias protects you from it.

2. Clever, educated people are less biased than others.

3. A group makes fewer biased judgements than one person alone.

4. Experienced professionals can trust their gut.

Bias in the wild

1

strongly agree

2

mostly agree

3

on the fence

4

mostly disagree

5

strongly disagree

5. Most biases are useful shortcuts, not mistakes.

6. Bias training at work is mostly a waste of money.

7. The best defence against bias is better procedures, not better people.

8. I am less biased than the average person in this lesson.

1. Is it fair to call a judgement "irrational" if it works most of the time?

2. Which institutions in your country are built to protect decisions from bias?

Private reflection

Task Choose one prompt and write privately. Nobody will read what you write. Sharing afterwards is optional.

Prompt 1

Think of a decision you, or someone you know, made with confidence and later regretted. Which of today's biases might have played a part, and what safeguard would have helped?

Prompt 2 (lighter)

Which of today's shortcuts do you think you use most when you shop, choose a film or pick a restaurant? Does it usually serve you well?

Before Lesson 6

This week, catch one bias in the wild.

Find one decision, advert or news story where an anchor, a vivid example or a stereotype did the thinking. Write it down in two sentences.

Next lesson: Numbers That Lie. Averages, percentages and charts that tell the truth and still mislead.

TEACHER'S REFERENCE · SKIP OR DELETE BEFORE CLASS

Answer key (1)

YOUR ESTIMATE / THE REVEAL (ANCHORING)

• Ethnologue's 2025 edition lists just over 7,000 living languages; other catalogues count differently, so any answer between about 6,500 and 8,000 is reasonable. What matters is the gap between the two group medians, not who was closest.

• For comparison, in Tversky and Kahneman's 1974 study, the group shown 10 gave a median estimate of 25% for African countries in the UN, and the group shown 65 gave 45%. (Today 54 of the 193 UN member states are African: about 28%.)

MEET LINDA

• A is more probable. Every feminist bank teller is a bank teller, so B can never be more probable than A. In the 1983 study, 85% of 142 University of British Columbia students chose B. When the same description was followed by eight statements to rank, the conjunction was ranked above "bank teller" by 89% of statistically naive students and by 85% of doctoral students in decision science at Stanford Business School.

WHAT COMES TO MIND?

1 Third. A typical English text contains about twice as many words with K in third position as words starting with K. In the 1973 study, participants' median estimate was 2 to 1 in the wrong direction for all five letters tested (K, L, N, R, V).

2 Roughly 40 to 1. WHO estimates about 300,000 drowning deaths in 2021; the Global Terrorism Index counted 7,555 deaths from terrorism in 2024. (The years differ, but the order of magnitude is stable.)

TWO SYSTEMS the ball costs 5 cents (the bat costs $1.05). The intuitive answer, 10 cents, would make the total $1.20. The puzzle is from Shane Frederick's Cognitive Reflection Test (2005).

TEACHER'S REFERENCE · SKIP OR DELETE BEFORE CLASS

Answer key (2)

POST-MORTEM suggested; most steps carry more than one bias

1 System 1 snap judgement and representativeness: Arjun matches Marta's stereotype of a head of engineering, and the decision comes before the evidence. Safeguard: agree the criteria and a scored, structured interview before meeting any candidate; no decision on the day.

2 Availability (one vivid failure retold at every meeting stands in for "quiet people fail"), plus judging from a sample of one. Safeguard: look at the record of all past hires, not the memorable one; judge internal candidates on the same scored criteria.

3 Anchoring: 210,000 euros becomes the reference point, so 185,000 euros feels like a win though it is far above the band. Safeguard: fix a salary range from independent benchmarks before any number is heard, and negotiate from the company's figure.

4 Confirmation bias (she asks only about successes) and motivated reasoning (an inconvenient warning is explained away as jealousy). Safeguard: give one board member the formal job of arguing against the hire; ask every candidate the same question about a failure.

5 Confirmation bias: she hears and records only the half of the reference that fits. Safeguard: a second interviewer takes notes; ask referees for specific examples, and with the candidate's consent speak to someone he did not choose.

6 Motivated reasoning (and, if groups raise it, sunk cost and public commitment): she has said publicly that it was her best decision, so the resignations must be the engineers' fault. Safeguard: exit interviews run by someone other than Marta; a pre-agreed review of every senior hire at three months.

• Step 7 is the outcome, not a decision. A good point to raise: judging a decision only by its outcome is itself a bias (outcome bias). Was the process bad, whatever happened?

TEACHER'S REFERENCE · SKIP OR DELETE BEFORE CLASS

Answer key (3)

• A pre-mortem (Gary Klein, 2007) fits the whole case: before deciding, the team imagines the hire has failed and lists why.

BIAS IN THE WILD evidence to offer after discussion; none of these settles the question

1 Weak protection on its own. Pronin, Lin and Ross (2002) found people readily see bias in others but much less in themselves (the bias blind spot). But structured training with feedback has reduced some biases for at least two months (Morewedge et al., 2015).

2 West, Meserve and Stanovich (2012) found that cognitive ability did not reduce the bias blind spot; if anything, higher ability went with a larger one.

3 Sometimes. Letting people consult each other sharply reduced errors on the Linda problem (Charness, Karni and Levin, 2010). But groups can also amplify bias: Lesson 10 returns to this.

4 Kahneman and Klein (2009) agreed that expert intuition can be trusted only in environments regular enough to learn, with an opportunity to learn them through practice and feedback (chess, firefighting), much less in poorly predictable ones (long-range political or market forecasting).

5 Gigerenzer's side of the argument (ecological rationality); the heuristics-and-biases tradition stresses the errors. Both sides accept that heuristics are usually useful and sometimes systematically wrong.

6 and 7. Open questions. The Morewedge result is one argument against 6; the Post-mortem safeguards are arguments for 7.

8 Not everyone can be less biased than average. If most of the group rated it 1 or 2, the group has just shown the bias blind spot.