Lecture: Bias and Decision-Making
Lecture script · about 15–20 minutes spoken · Week 14
Opening
Think about the last time you looked something up to win an argument. You typed in your side, read the first few results, and felt better. Now picture a jacket with a tag that reads "was 200 dollars, now 120". The 120 feels like a deal before you've asked what the jacket is worth. And then there's the group project your team swore would take 6 days, and that took 9.
A nursing student meets this when a first impression of a chart starts steering every later reading of it. A business major meets it when the first salary figure named in an interview shapes everything after. A shopper meets it at every "was" price. In each case, the slip isn't in someone else's argument. It's in your own judgment.
Here's where we are. Week 2 gave you standard form and the principle of charity. Week 12 taught you to check a source, and Week 13 taught you to ask what would show a claim wrong. For most of this course you've tested other people's arguments. In this lecture I'll turn those tests around and point them at you, in seven steps, ending with one bias check you can run on your own reasoning.
1. Catch confirmation bias
Picture a friend who's sure the campus coffee cart is overpriced. Every time they see a high price, they point it out. Every time they see a fair one, they say nothing. After a month they have a long list of proof, and all of it fits what they believed on day one.
That pattern has a name. A cognitive bias is a steady tilt in how people gather or weigh evidence that pulls their judgments the same wrong way. The best known is confirmation bias, the habit of seeking and favoring evidence that fits what one already believes, while overlooking or discounting evidence that could show it wrong.
The classic experiment comes from Wason, in 1960, in a paper called "On the failure to eliminate hypotheses in a conceptual task", in the Quarterly Journal of Experimental Psychology. One honest caveat: it's the classic experiment later read as showing confirmation bias. It isn't the paper that coined the phrase.
Here's how Wason's rule-discovery task worked. Participants were told that the numbers 2, 4, 6 fit a rule. Their job was to find the rule by proposing sets of three numbers of their own, and each time they heard whether their set fit. The rule was any ascending sequence, each number larger than the one before. Most people tested sets that fit their own guess, rather than sets that could prove the guess wrong.
They weren't careless. They were trying hard. That's what makes this bias worth your attention: it doesn't feel like a mistake from the inside.
Pause and try this. Your friend with the coffee-cart list says confirmation bias is something stubborn people have. Wason's participants were trying to find the rule. What does that tell you?
Think about who was in the experiment.
Here's the answer. Confirmation bias isn't only other people's problem. Careful, motivated people showed it too, so the safe move is to assume the tilt is in you and check for it on purpose.
2. Run the disconfirming test
Let's watch one player go through Wason's task. The player sees 2, 4, 6 and guesses "add 2 each time". Then the player tests three sets that fit the guess: 8, 10, 12, then 20, 22, 24, then 1, 3, 5. All three get a yes. The player announces "add 2", and it's wrong.
Why didn't the tests catch it? Every set that fits "add 2" also fits "any ascending sequence". So none of those three tests could have shown the guess wrong. Each one was a confirming test, a test chosen because it should come out in a guess's favor, so that a pass adds little.
Now the player tries 1, 5, 20. That set breaks the guess. If "add 2" were the rule, it would get a no. It gets a yes, so "add 2" is wrong. Then the player tries 6, 4, 2, and gets a no, which points toward the real rule. That's a disconfirming test, a test that could show a guess or belief to be wrong if it is wrong.
This is Week 13's logic of a test, turned on your own guess. A prediction that comes true supports a guess without proving it. A prediction that fails counts against it. So aim your test where your guess would break.
Pause and try this. The player's three confirming tests all got a yes. The one disconfirming test also got a yes. Which taught the player more, and why?
Picture what each test could have shown.
Here's the answer. The single disconfirming test taught more. One test that could fail beat three that couldn't, because only that set could tell "add 2" apart from the real rule.
3. Audit your own evidence
Now let's bring this home to the running question: should the campus library stay open around the clock during finals week? Say you lean yes, and you open a search engine.
Here's a student who did exactly that. They used 3 search phrases, each asking about the benefits of late-night study, and found 10 sources. Of the 10, 8 favor overnight hours and 2 support the objection. It's tempting to read that as the evidence leaning yes. But all 3 search phrases pointed one way. So the lean of 8 to 2 tells you more about the search than about the question.
What the student did here is called an evidence audit: listing the evidence one gathered on a question, with the searches that found it, and checking which side it favors and whether the search itself produced the lean.
The first repair is to consider the opposite, the habit of stating the strongest case against one's own view, fairly and in full. You already own the tool for that. Put the objection in standard form, with charity. Premise O1: overnight hours encourage all-night studying. Premise O2: losing sleep can hurt exam performance. So, the conclusion OC: keeping the library open overnight would help students harm themselves. On the map, that's an arrow marked "against", pointing at the running question's conclusion.
The second repair is a search that could go against you. The student runs 2 more phrases about the harms of all-night studying and finds 10 more sources. Of those, 3 favor overnight hours and 7 support the objection. The full list is now 20 sources, 11 on one side and 9 on the other.
Pause and try this. The full list has 11 sources favoring overnight hours and 9 supporting the objection. Does the side with 11 win?
Think about what a count can and can't tell you.
Here's the answer. No. A count of sources isn't a verdict. One strong study can outweigh several posts that copy one story, so the student now weighs each source with the source check.
4. Spot motivated reasoning and overconfidence
A fan watches a referee's call against their team in slow motion and finds reason after reason it was wrong. The same fan sees an identical call against the rival team and accepts it at full speed. Same kind of evidence, different standard.
That's motivated reasoning: evaluating evidence more leniently when it supports a conclusion one wants and more strictly when it does not. The cure is the same-standard test: would I accept this evidence if it pointed the other way?
Here's a case. Two posts each rest on one student's story. One says overnight study helped a student's grades. The other says an all-nighter hurt a student's exam. A student who favors overnight hours accepts the first and dismisses the second as "only one story". The complaint is fair, and it applies to both posts. Both are single anecdotes, so the same-standard test gives both the same low weight.
The other tilt in this section is overconfidence: holding a belief or a prediction with more certainty than the evidence supports. A team of 4 says it's sure a project will take 6 days. Its last 3 projects took 9, 10, and 8 days. Add those up and you get 27, and 27 divided by 3 is 9 days on average. That's 3 days longer than the plan, and every one of the 3 past projects ran past 6 days.
Pause and try this. The team can define overconfidence perfectly. Does knowing the word protect them from planning 6 days?
Think about where the 6 came from.
Here's the answer. No. Knowing about a bias doesn't protect you from it. The team needed a habit, checking the record before committing, and the record pointed to 9 days.
5. Name the shortcuts: availability and anchoring
You guess how many people in your building own a bike by picturing the bikes in the hallway. Most days that works. Shortcuts like that are everywhere, and most of the time they serve you well.
A heuristic is a mental shortcut for making a judgment quickly, which often works and sometimes misleads. Tversky and Kahneman, in 1974, in an article called "Judgment under Uncertainty: Heuristics and Biases" in the journal Science, described availability and anchoring, along with a third heuristic, representativeness. We'll take the first two.
The availability heuristic is judging how common or likely something is by how easily examples of it come to mind. After a news story about a plane crash, a traveler decides flying is too dangerous and drives instead. The crash is vivid and easy to recall, and ease of recall isn't frequency. The better question is how many crashes happen out of how many flights, a base rate the story doesn't give.
Anchoring is the pull of a starting number on an estimate, which keeps the estimate too close to that number even when the number is arbitrary. Back to the jacket. The tag says "was 200 dollars, now 120". The discount is 80 dollars, which is 40 percent off. That arithmetic is right. But the 200 is the anchor, and whether 120 dollars is fair is a separate question. You answer it by comparing prices elsewhere.
Here's a classroom version with illustrative numbers. A class of 40 is split into two groups of 20. One group is asked whether a used bike costs more or less than 300 dollars, the other whether it costs more or less than 30 dollars. Then everyone estimates. The first group averages 180 dollars, the second 90 dollars. That's a 90-dollar gap from a number nobody claimed was the price.
Pause and try this. Is a heuristic always a mistake?
Think about the bikes in the hallway.
Here's the answer. No. A heuristic is a shortcut that often works. It misleads in particular conditions, like a vivid story or an arbitrary number, and that's when you slow down.
6. Practice the four debiasing habits
A carpenter measures twice before cutting once. The habit doesn't make the carpenter smarter. It catches errors the carpenter would otherwise make.
The debiasing habits work the same way. There are four. Consider the opposite. Look for the case that would disprove your view. Write your prediction down before you see the result. Decide in advance what evidence would change your mind.
Watch the student from the running question use all four. They considered the opposite by writing O1, O2, and OC. They looked for the disconfirming case with 2 search phrases aimed at the other side. Before reading the 10 new sources, they wrote down a guess about which way those sources would lean. And they wrote what would change their mind: evidence that the library's sleep reminders and quiet rest floor don't reduce all-night studying would give the objection more weight.
Notice that the student still favors overnight hours. The habits don't erase your view. They make sure your view has met the strongest case against it.
Pause and try this. Of the four habits, why does deciding in advance what would change your mind come before the evidence arrives?
Think about what happens to a standard set afterward.
Here's the answer. A standard written before the evidence can't bend to fit it. Set afterward, it tends to move wherever the evidence you like happens to land.
7. Run the whole bias check
A pilot runs the same checklist before every flight. The list is short, and it works because it doesn't depend on feeling careful that day.
Here's yours. The bias check asks five questions of your own reasoning. What did I search for, and which side did it favor? Would I accept this evidence if it pointed the other way? What number or vivid story is steering me? What did I predict before I looked? And what would change my mind?
Run it on the week. The player's 3 confirming tests couldn't catch the wrong guess, and 1 disconfirming test did. The student's 8 to 2 lean came from the search phrases, and a search for the other side changed the list to 11 and 9. The posts, the plane crash, the price tag, the bike estimates, and the project plan each show a bias steering a judgment. And none of these checks decides the running question.
Pause and try this. Suppose you find that someone reached their view through confirmation bias. Does that show their view is false?
Think about a biased route and a true destination.
Here's the answer. No. A biased search can still land on a true conclusion. Finding the bias tells you to recheck the evidence fairly, not which way the recheck will come out.
Closing
Let's pull it together. Confirmation bias is seeking and favoring evidence that fits what you already believe, and Wason's task shows it in careful people. A disconfirming test could show your guess wrong, so it teaches more than many confirming tests. An evidence audit asks whether your search made the lean, and considering the opposite puts the objection in standard form. The same-standard test catches motivated reasoning, and a track record catches overconfidence. Availability and anchoring are shortcuts that often work and sometimes mislead. The four habits are things you do, and the bias check runs them in order.
Next in this module come the slide deck, the Readings and Resources page, the lecture tutorial, the practice set, and the quiz. Next week, in Reasoning About Values and the Course Synthesis, considering the opposite grows into stating the other side at its strongest.
Built with The Course Maker · samples.thecoursemaker.com