Chapter 13: Decision Making, Biases & Cognitive Illusions

Welcome to Last Minute Lecture.

This free chapter overview is designed to help students review and understand key concepts.

These summaries supplement not replaced the original textbook and may not be redistributed or resold.

For complete coverage, always consult the official text.

Welcome back to The Deep Dive, the show where we take a stack of your sources, articles, research, complex concepts, and extract the critical insights you need to be instantly well -informed.

Today we are tackling a big one.

A huge one.

We are cracking open one of the most fundamental and I think maybe one of the most error -prone processes of human cognition, decision -making.

It really is.

I mean, it's the cognitive activity that underpins nearly every significant action we take.

We're diving into how the human mind chooses among alternatives which, and this is the critical part, often happens under conditions of uncertainty.

Right.

Think about the high -stakes decisions you make.

Choosing where to live, whom to marry, or even you know what product to invest in.

These choices are almost never made with perfect information.

Exactly.

We're not talking about choosing between a red apple and a green apple here.

We're focused on that moment when you, our listener, are staring down a major life change, like a career shift, or picking a college major.

You're feeling agitated, maybe a little nervous, and you have this feeling of just severe information overload.

Why?

Because you're juggling these conflicting goals.

Maybe you want a high salary and job satisfaction and low travel.

Right.

They all pull in different directions.

And you face the staggering number of potential options and criteria to evaluate.

So our mission today is to really unpack the strategies people use to manage that torrent of information, uncover where they systematically stumble, and maybe reveal how we can achieve better, more rational outcomes.

And we really have to start by defining what we mean by rationality here.

Because the standard for success isn't just individual success.

I mean, luck plays way too big a role in whether any single decision turns out well.

Oh, for sure.

The real yardstick, the one put forth by cognitive psychologists like von Winterfeld and Edwards, is rational decision making.

And to be rational, it just means selecting ways of thinking and acting that serve all of your relevant goals and principles.

So it's holistic.

It's not just one thing.

If I'm trying to decide on a new laptop, I can't just pick the sleekest one or the cheapest one.

I have to balance speed and battery life, memory, software compatibility.

If I ignore speed because I'm totally fixated on price, I've basically violated my own goal of having a functional work tool.

Precisely.

And rationality also puts a requirement on your information gathering process.

It has to be painstaking and fair.

It demands that you actively search for evidence that might contradict what you already think.

That's the hard part.

It's the hardest part.

If you want that sleek, cheap laptop, but the research shows it consistently fails after a year,

well, rationality demands you weigh that contradictory evidence fully.

No ignoring it.

That makes perfect sense.

But if the ideal is rationality, where do we consistently fall short?

Because let's be honest, we all make decisions based on emotion or gut feelings that we later regret.

The primary culprit, and this really sets the stage for every cognitive error we're going to talk about today is cognitive overload.

We just have far too much available information for our limited cognitive processing capacity to handle effectively.

When the system gets overwhelmed,

just try to imagine holding 20 different criteria in your head at the same time.

Impossible.

It's impossible.

So we're forced to rely on coping strategies.

These strategies, which we call mental shortcuts or heuristics, are brilliant at saving time and energy, but they're also the pathways to systematic error and irrationality.

We use them because we have to.

We just don't have infinite mental bandwidth.

Okay.

So let's unpack this journey of decision making because it's not one single moment, right?

It's a whole process that psychological research breaks down into, I think,

five distinct tasks or phases.

And as you mentioned, this process is rarely linear in real life.

If you were to look at a conceptual flow chart, like the one in figure 13 to one in the literature, you'd see arrows looping back all over the place.

So you might gather some information.

And then realize you need to redefine your goals, or you might make a choice and then immediately start evaluating whether you chose wisely, which forces you to go back and reevaluate the options all over again.

That non -linearity is a key insight.

Let's detail those five phases, starting with the very first one, which really defines the entire decision space, setting goals.

Correct.

Before you even look at your alternatives, you have to establish what you're trying to accomplish.

And this involves a really critical assessment of your future plans, your core values, your principles, and your overall priorities.

Why are you even making this decision?

So a student, for instance, might set the goal of choosing biology because their ultimate aspiration is medical school.

Which requires a specific set of foundational knowledge.

Exactly.

Conversely, another student might choose economics because their goal is getting into a competitive corporate training program,

which requires a totally different set of skills and academic history.

The goals fundamentally determine which paths you even start to consider.

And once those goals are clear, we move into what feels like the heavy lifting, gathering information.

This is the data collection phase, and it requires identifying all the viable options, researching the likely short -term and long -term consequences of each one, figuring out who else might be affected by your choice.

And critically, right, assessing whether an option opens or closes off future paths.

That's a huge one.

If you're buying that computer, you research specific models, customizations, features.

But you also have to gather information about the criteria themselves.

Like, what even constitutes adequate speed for what I need to do?

How important is brand reliability, really?

This task can get overwhelming very, very quickly.

Which brings us to the third phase, which is absolutely critical for complex choices, structuring the decision.

This is where we try to, like, tame that torrent of information.

Exactly.

When you have multiple options and multiple criteria, the data just becomes staggering.

Let's go back to the college major example.

Studies show that students considering a major often look at about four different majors and about seven different criteria, things like their interest, the difficulty, future earning potential, faculty quality.

So if you just multiply options four by criteria seven, that's 28 distinct pieces of information, 28 little cells in a mental spreadsheet you need to evaluate and weigh.

Right.

How much do I enjoy chemistry versus how relevant is psychology to my career goals?

Decision structuring is simply the way the decision maker organizes this staggering amount of data to create a coherent,

manageable framework for comparison.

It's the mental effort of building that spreadsheet when your brain would much rather just pick the major that sounds coolest.

That mental heavy lifting, organizing everything into a coherent framework is so essential.

But assuming we structured the data, what are the final two phases?

The fourth phase is making a final choice.

This is the moment of selection.

For simple decisions, it's almost automatic.

For complex ones, though, it often involves a preceding meta decision.

A decision about the decision.

Exactly.

For example, deciding when to stop gathering information.

Have I researched enough?

Or deciding which specific subset of the information I gathered is now the most relevant.

And the final phase, which you noted is critically important, yet so often just skipped, is evaluating.

Yes.

Evaluating is the cognitive feedback loop.

It's the process of reflecting on the decision process itself.

Not just, hey, was the outcome good, but - Did I structure the data correctly?

Did I weight my criteria accurately?

What went well in my strategy and what parts of my decision -making approach should I revise for the future?

The failure to evaluate prevents any systematic improvement in our cognitive decision skills.

If you don't evaluate your method, you're just doomed to make the same errors next time.

That's a profound point.

We stop when we feel relieved, not when we've learned something.

Okay, let's transition.

Because the information gathering and structuring phases rely so fundamentally on our ability to assess risk and opportunity.

They require us to understand, and often guess at, probability.

Since almost all important decisions happen under uncertainty, we are constantly engaged in chance estimation.

Probability is the mathematical measurement of a degree of uncertainty expressed as a number between zero and one.

Zero means certain not to happen.

One means certain to happen.

Right.

Conceptually simple.

But research shows that the average person, someone without statistical training, struggles enormously with anything other than those extremes.

We are ill -equipped to deal with intermediate values, like .3 or .7.

We struggle to intuitively grasp the real difference between saying we're 30 % sure and 40 % sure in a high -stakes, real -life context.

And this is where our deep dive into cognitive illusions really begins, showing how easily our intuition leads us astray.

Especially when dealing with something called base rates.

We absolutely have to walk through the classic mammogram problem.

Yes, the one from Barron.

This example demonstrates so perfectly how poorly our minds handle conditional probability.

It's a fantastic illustration.

Okay, let's lay out the scenario carefully for everyone listening.

A 30 -year -old woman finds a lump, goes to her physician.

The physician knows that in the general population of women her age and health profile, only about five in a hundred actually have breast cancer.

And that is the crucial low base rate.

Five percent, that's the starting point.

Right.

Now, a mammogram is administered.

The test itself is pretty good.

It indicates cancer 80 % of the time in women who do have cancer.

That's the true positive rate.

But critically, it also produces a false positive.

It indicates cancer in healthy patients 20 % of the time.

The woman gets a positive test result.

So she gets a positive reading.

Now, if you are listening, pause for a second and ask yourself, what is the probability that she actually has cancer?

Given the test is 80 % accurate, most people, their intuition just screams an estimate of 50%, 60%, maybe even 80%.

Because that 80 % accuracy figure is so salient.

It's so available to your memory.

But that is the cockative trend.

That is the real answer.

The calculated answer, which requires using Bayes' theorem, is only 0 .1717%.

The intuitive estimate is often four or five times higher than the actual probability.

That is genuinely shocking.

We need to explain how the calculation gets us to 17 % without using complex equations because the error is all about how we mentally process that initial base rate.

Let's scale it to 100 random women.

Makes it much clearer.

So based on that base rate, we expect five women out of the 100 to have cancer and the other 95 women to be healthy.

Okay.

So start with the true cases, the five women who have it.

Of those five women with cancer, the test accurately detects 80 % of them.

So five times 0 .80 is four.

Four women get a true positive result.

These are the truly sick women who test positive.

Okay.

That's straightforward.

Now the false positives.

This is where that base rate really plays its devastating role.

We have 95 healthy women.

The test falsely indicates cancer in 20 % of them.

20 % of 95 is 19.

19 healthy women receive a false positive result.

I see it now.

The light bulb just went on.

So the total number of people who received a positive test result is the four true positives plus the 19 false positives, which is 23 positive results in total.

Precisely.

And of those 23 people who got that positive result, only four of them actually had cancer.

So the probability that our woman actually has cancer given her positive test is four divided by 23, which is about 0 .17, 17%.

Wow.

That extremely low initial base rate, the fact that only 5 % of this group had cancer to begin with, is the essential piece of information that our intuition just neglects.

And that leads to that massive overestimation.

That example perfectly highlights the difference between the kinds of probabilities we use.

The mammogram calculation uses objective probability, which is a statistical measurement not influenced by the estimator's personal characteristics.

But in real life, when I estimate that a project will succeed because I'm feeling optimistic, I'm using subjective probability.

And subjective probabilities are influenced by all those personal factors.

Your mood, your optimism, your personal experience with similar ventures.

In many real life circumstances, like trying to predict the stock market or deciding if your new business idea will work,

objective probabilities simply aren't available.

We're forced to rely on those subjective estimates, making us prone to the very large errors we just discussed.

Okay, that sets the stage perfectly.

Now we understand why we fall short cognitive overload and how we miscalculate chance base rate neglect.

So let's dive into the core systematic errors.

When people gather and assess information and deviate systematically from probability or logic, these errors are labeled cognitive illusions.

The term is designed to be analogous to perceptual illusions.

Just as we see lines of equal length as different because of the arrows on the end, our cognitive system produces these predictable errors when we use mental shortcuts or heuristics to cope with complexity.

So these illusions aren't evidence that the whole system is faulty.

Not at all.

They're just instances where our normal efficient processing mechanisms lead to a predictable mismatch with rational reality.

And we are going to dive deep into these heuristics, starting with one of the most famous availability.

The availability heuristic is the rule of thumb that causes us to assess the frequency or probability of an event based on the ease with which we can retrieve, construct, or associate instances of that event in our minds.

If something comes to mind quickly, we assume it must be frequent or likely.

A classic demonstration of this from Tversky and Kahneman involves the letter L.

So is L more likely to appear in the first position of a word or the third position?

Almost everyone guesses the first position.

And why?

Because it is infinitely easier to search our mental lexicon, our internal mental dictionary, for words starting with L, like library or lament.

We organize our mental dictionary by initial letters.

But trying to find words with L in the third position, like bell or sail, requires a much more difficult and less direct search process.

Right.

So the ease of retrieval for initial letters makes that position more available to us, leading us to overestimate its frequency when, statistically, L actually occurs more frequently in the third position.

Another brilliant, totally counter -intuitive example is the committee problem.

If you have 10 students and you're asked to estimate the number of distinct two -person committees you could form versus the number of distinct eight -person committees.

Intuition almost always says there are way more two -person committees.

But mathematically, the numbers exactly equal 45 of each.

The reason two -person committees seem more frequent is that they're inherently more distinct.

They share less membership, so they're easier to visualize or construct mentally than these huge eight -person committees which have massive membership overlap.

That distinctiveness increases their availability, making us judge them as more numerous.

And we see this availability heuristic play out in real life in truly interesting ways, particularly in shared relationships.

Ross and Sickly did a study where they surveyed married couples.

They asked them separately to estimate their own contribution to 20 different household activities, from making breakfast to doing the yard work.

And the finding was systematic.

Both husbands and wives consistently overestimated their own contribution to the majority of those activities.

And why?

Because your own efforts, your own thoughts about starting a task, your planning, your execution, they're all more available and salient to your memory than the partner.

You are always present for your own actions.

You might miss theirs entirely.

Exactly.

When you're asked to recall examples of contributions, you simply retrieve more of your own actions.

And that cements the perception that you carry the greater share of the burden.

Which means the availability heuristic can lead directly to conflict.

Okay, let's move on to the second major heuristic, which often clashes with those base rates we talked about.

Representativeness.

The representativeness heuristic is the assumption that causes us to expect two things.

First, we expect a random process to produce results that look random.

And second, we expect small samples to resemble the large population from which they are drawn.

This is why if you flip a coin six times and get heads, heads, heads, tails, tails, tails, and then you flip another six times and get tails, heads, tails, heads, heads, tails,

we instinctively feel like the second sequence is more likely.

Even though both are equally likely.

But the first sequence,

three straight heads, three straight tails, it just doesn't look representative of a random process.

The second sequence, with its jumbled up results,

fits our mental prototype of randomness.

And this is the cognitive foundation of the gambler's fallacy.

Oh, absolutely.

If a roulette wheel lands on black eight times in a row, our internal expectation of randomness makes us believe that the next spin must be read to even out the sequence.

But the roulette wheel has no memory.

The trials are independent.

Completely independent.

This mistaken belief is tied to the assumption of what's called the law of small numbers.

People expect small samples, like those eight roulette spins or maybe the five people they talk to about a new car, to show the same proportions and distributions as a population of 10 ,000.

But in reality, small samples are highly unreliable, and they're far more likely to deviate wildly from the population average.

Yet we rely on them because they are available and vivid.

And the power of representativeness is most clearly shown when it completely bulldozes rational statistical data.

This was demonstrated in the classic Tom W.

study.

Right.

So participants were given a detailed personality sketch of a fictional graduate student, Tom W.

He was described as intelligent, but a bit rigid with a need for order and

little sympathy for others.

And they were asked to predict the likelihood that he was studying in one of nine different graduate specializations, like computer science, law, or humanities.

But what they should have done is incorporate the base rates.

A separate group was first asked to estimate the actual percentage of all graduate students in the US currently enrolled in those nine fields.

And the base rates showed that fields like humanities and education had vastly higher enrollment numbers than computer science or engineering.

But the participants in the prediction group, they largely ignored those base rates.

When they were asked to predict Tom W.'s field, their likelihood rankings matched a third group's rankings of how similar Tom W.'s personality was to the stereotype of a student in that field.

So Tom W.

sounded like a classic stereotype of an engineering or computer science student, the rigid, orderly type.

So people ranked those fields highest, completely ignoring the fact that far more students overall were in humanities.

The similarity to the stereotype of representativeness just trumped all the life through the use of man who arguments.

A huge statistical study might show conclusively across 10 ,000 patients that a certain vaccine is safe and effective.

And yet someone will dismiss that rigorous data instantly with, well, I know a man who got that vaccine and then got sick three days later.

That single vivid case, a sample size of N1, is so representative of a negative outcome and so available and compelling as a story that it's used to dismiss massive statistically reliable data sets.

Nisbet and Ross called this a clear failure of using the law of small numbers.

It's a powerful, powerful bias.

Let's shift our focus now to how the subtle context of a decision can dramatically alter our choice, even if the outcomes are logically and, you know, monetarily identical.

We are now talking about framing effects.

Framing effects demonstrate that we don't evaluate outcomes based on their absolute value.

Instead, we evaluate them as gains or losses relative to a reference point, which is usually our current state.

So how the situation is described or framed establishes that reference point and therefore dictates our decision, often irrationally.

The classic gas station example is perfect for illustrating this.

You've got two stations, identical prices.

Station A charges $2 a gallon, but advertises a five cents gallon discount for cash.

Station B charges $1 .95 a gallon, but announces a five cents gallon surcharge for credit cards.

And in both cases, the cash price is $1 .95, and the credit price is $2.

The economics are identical.

Yet the vast majority of consumers prefer Station A.

They prefer the discount because they view it as a gain.

You're saving a nickel from that $2 reference point.

Station B frames the situation as a loss.

You start at the lower price, and then you're charged or you lose a nickel if you use your card.

And Kahneman and Tversky's work on this showed a consistent finding.

We treat losses more seriously than equivalent gains.

We are loss averse.

The pain of losing a dollar is psychologically greater than the pleasure of gaining a dollar.

So by changing the linguistic frame, you change the psychological reference point, which in turn changes whether we see the transaction as risk taking or security seeking.

That emotional weighting of loss versus gain explains so much of our irrational economic behavior.

Okay, moving from framing to numbers, let's discuss anchoring.

This is where an initial, often completely irrelevant, starting value disproportionately influences a final numerical estimate.

Anchoring is a deeply reliable cognitive illusion.

Researchers showed this by asking participants to estimate the population of Philadelphia.

But before they answered, they watched the experimenter spin a roulette wheel that generated an arbitrary random starting value.

And even though participants knew that the number was random and totally irrelevant to Philadelphia's census data, the initial numbers still pulled their final estimates.

Those who saw a high number on the wheel, say 5 million, gave significantly higher final estimates for Philadelphia's population than those who saw a low starting number.

That initial irrelevant anchor exerted this gravitational pull on the subsequent cognitive adjustment.

And this anchoring effect works even when the starting numbers aren't arbitrary, but just the first ones you encounter in a calculation.

Right.

There was an experiment where participants had to estimate the value of 8 times 7 times 6, all the way down to 1, versus 1 times 2 times 3, all the way up to 8.

The correct answer is 40 ,320, and everybody grossly underestimated it.

But the group starting with 8 times 7 times 6 gave a much higher estimate than the group starting with 1 times 2 times 3.

The theory is that people do the first few multiplication steps, and then they just insufficiently extrapolate or adjust their estimate from that initial partial calculation.

The higher result of 8 times 7 times 6 anchors, their subsequent estimation is significantly higher than the lower result of 1 times 2 times 3.

The adjustment away from the anchor is almost always inadequate.

Next up is the surprisingly common irrationality that has ruined many a business venture or government project, the sunk cost effect.

The sunk cost effect, defined by Blumer, is the irrational tendency to continue funding or effort or commitment in a failing project simply because of the investment of money, time, or emotion you've already spent on it.

We see this when a company keeps pouring money into developing a product that clearly isn't selling just because they already invested $5 million in R &D.

Exactly.

The core error is confusing resources that are already spent, which are gone, they're sunk, with the expected future benefits and costs of continuing.

Rationally, past investment should be irrelevant.

The decision to continue should only be based on whether the expended benefits of the project moving forward outweigh the expected future costs.

But humans find it emotionally almost impossible to accept that they've wasted time or money, so they irrationally choose to double down on the failure, hoping to somehow redeem that sunk cost.

Moving on to another perceptual bias, we encounter illusory correlation.

This is the error of seeing a non -existent association between two variables just because a relationship seems plausible or expected based on a prior bias or stereotype.

A simple demonstration of this is the hair twisting example.

If you're presented with data showing a group of students categorized by whether they are under stress or not and whether they are hair twisters or not, and you actually calculate the data, you find the proportion of

25 % for both the stressed group and the non -stressed group.

There is no statistical relationship whatsoever.

Yet, when you ask people about it, they often report seeing a weak correlation.

Precisely.

Because hair twisting sounds like a nervous anxiety -related behavior and nervous behaviors are plausibly correlated with stress,

our pre -existing bias makes us impose that relationship.

We see the correlation even when the data says otherwise.

This was famously shown in clinical context with the whole draw -a -person test controversy.

Clinical psychologists were reporting correlations between specific drawing features like atypical eyes and specific symptoms like suspicious clients.

So, researchers like Chapman and Chapman gave undergraduates randomly paired drawings and symptoms, ensuring there was no real relationship in the data set they were given.

And yet, the undergraduates still discovered the same expected correlations, atypical eyes paired with suspicious clients.

It just shows how powerfully our prior associations and expectations color our judgment, causing us to impose relationships that are simply not present in the evidence.

Next is a bias that makes us unreliable witnesses to our own history.

Hindsight bias.

This is the tendency to consistently exaggerate what we could have anticipated in foresight after an event has already happened.

The classic hindsight is 2020.

If your friend tells you they knew all along that the political outcome would turn out the way it did or that you were going to change your major, they're probably suffering from hindsight bias.

Because once the outcome is known, it becomes incredibly easy to construct a narrative or find evidence that makes that outcome seem inevitable.

Right.

Their actual ability to predict it beforehand was likely far weaker.

Fischhoff demonstrated this by having participants read historical passages like the details of the British war with the Gurkhas in Nepal and then having them estimate the likelihood of various outcomes.

The critical manipulation was telling some participants what the actual outcome was versus not telling others.

And the participants who were told the outcome say a British victory subsequently gave significantly higher estimates that that outcome would have happened compared to those who were kept in the dark.

Once the result is known, we look back, we rearrange our understanding, and we make the past seem entirely predictable.

This is a crucial failing when we evaluate decision makers because we inevitably look at their choices through the distorted lens of hindsight.

A related bias linking our tendency to manage past and future evidence is confirmation bias.

Confirmation bias is the tendency to actively seek information that confirms an initial hypothesis and often just as powerfully to ignore or downplay contradictory evidence.

So if you're deciding whether or not to adopt a new marketing strategy and you already kind of lean towards doing it, you might only seek out case studies where similar strategies succeeded.

Right, and that seems human, but it's not rational if your goal is an unbiased assessment.

If you are truly seeking the best strategy, you need to talk to a random sample or specifically seek out those who tried the strategy and failed or who chose a different strategy and succeeded.

By only searching for information that confirms your initial hunch, you prevent yourself from getting the balanced view necessary for a rational choice.

Rationality demands you actively seek disconfirmation.

It's a tough discipline.

All these specific biases availability,

representativeness, anchoring, framing, they all contribute to the most general bias of all.

Overconfidence.

Overconfidence is a pervasive and profound impediment to better decision making.

To measure it rigorously, researchers use what are called calibration studies.

They give people a long list of general knowledge trivia questions and they ask them not only to answer but also to rate their subjective confidence in their answer from 0 .5, which is a pure guess, up to 1 .0, which is 100 % certain.

Then they compare the subjective confidence rating with the actual accuracy over many, many trials and they generate a calibration curve.

So if you were plotting this on a graph.

The ideal perfectly calibrated decision maker would have their curve plot exactly on the 45 degree line.

If they say they are 80 % confident in a set of answers, they should be correct 80 % of the time.

But that's not what happens, is it?

Not at all.

The reality is that typical calibration curves are consistently bowed out from that ideal line.

For example, for all the questions a participant rates is 80 % confident, they might only be correct 60 % of the time.

Even when participants claim 100 % certainty, their actual accuracy rarely exceeds 75 % to 80%.

So we are systematically convinced we are better estimators than we actually are.

Our impressions of our own judgment are just consistently inflated.

What's the implication of that?

Overconfidence acts as an arrogance in decision making.

If you genuinely believe your judgment is already highly accurate, that you're hitting 80 % when you're actually only hitting 60%, you won't see any need to critically examine your process or seek decision aids or accept guidance.

Overconfidence makes improvement basically impossible.

Okay, we've established how we fall short using our heuristic driven minds.

Now let's pivot to the formal models developed to show us how we should integrate information.

We are moving into the utility models of decision making.

And this is where we need to introduce a critical distinction between different types of models.

Normative models define ideal performance, how a theoretically perfect rational agent should decide.

Okay, the ideal.

Then you have prescriptive models, which suggest how we ought to decide, taking real world limitations and human cognitive limits into account.

And finally, descriptive models, which simply detail what people actually do in real life scenarios.

Our starting point here is expected utility, EU theory, which is a quintessential normative model.

It's an evolution of the older, simpler, expected value theory, which only focused on money.

Right.

EU theory goes beyond mere dollars and cents to focus on utility, which captures abstract concepts like happiness, satisfaction, personal fulfillment, and achieving long -term goals.

The fundamental goal of EU theory is to maximize expected utility.

The choice with the highest EU is the rational choice.

And this is captured by the formula.

EU equals the sum of the probability of each outcome times its utility.

Exactly.

You multiply the subjective probability of each potential outcome by its utility, the satisfaction you get from it, and then you sum those values across all possible outcomes.

Let's walk through the major selection example from table 1301 in literature to make this concrete.

Say you are comparing psychology and chemistry as majors.

You've subjectively determined the utility or the satisfaction value of success and failure in each.

Success in psychology has a utility of 35 for you.

Failure, a utility of negative 20.

And you estimate your probability of success in psychology to be 0 .60.

Okay, so for psychology, the EU calculation is probability of success, 0 .60 times its utility, 35, plus the probability of failure, 0 .40 times its utility, negative 20.

And that comes out to 21 plus a negative 8, which is 13.

So the expected utility for psychology is 13.

Right.

Now let's look at chemistry.

Maybe you're less confident here, so you estimate the probability of success lower at 0 .45.

You subjectively value success at 30 and failure at only 4 or less negatively, because maybe chemistry is easier to pivot from.

Okay, so the chemistry EU calculation would be 0 .45 times 30 plus 0 .55 times 4.

Which yields 13 .5 plus 2 .2, which is 15 .70.

So even though you subjectively value the outcome of succeeding in psychology higher at 35 utility points than chemistry, which was 30, the expected utility calculation dictates that choosing chemistry at 15 .70 is the more rational choice than psychology at 13.

Because it's about maximizing overall satisfaction over time.

This is why EU theory is the normative model.

It's the ideal standard of choice.

But as you pointed out earlier, major life decisions are rarely about a single dimension, like success or failure.

We often care about seven or eight dimensions simultaneously.

Difficulty, appeal, career applicability, work -life balance.

And that complexity makes cramming everything into one single utility number pretty much impossible for the average person.

Which brings us to multi -attribute utility theory, or MAUT.

This is designed specifically for these kinds of complex decisions with multiple potentially conflicting goals.

But it's still a normative model, right?

Because it demands that same systematic rigor.

It does.

MAUT requires six structured steps to analyze the choice.

Step one, break the decision into independent dimensions.

For a job offer, these might be salary, commute time responsibilities, and corporate culture.

And you have to make sure these dimensions are truly independent.

Step two is the most psychologically challenging for the decision maker.

Determine the relative weights of each dimension.

You, the decision maker, have to decide how important each factor is to you relative to all the others.

And that's where the hard work is.

You might weight salary at 30 % importance, culture at 40%, and commute time at 30%.

This weighting is entirely subjective and highly variable between individuals, as you can see in diagrams like figure 13 to 3.

We might say we value work -life balance most.

But when forced to assign a numerical weight, we might find ourselves putting a higher number on salary, revealing our true cognitive calculus.

Okay, so then step three is just listing all the alternatives, all the job offers.

Step four is ranking each alternative along all five dimensions.

Step five is multiplying the ranking of each alternative by the weighting of that dimension.

And then step six is?

Choose the alternative with the highest final weighted value.

And the MAUT process inherently requires a compensatory strategy.

That means a high score on one dimension, say incredibly high job satisfaction, can compensate for a low score on another dimension, like a slightly lower salary.

You are making explicit trade -offs.

So if I accept JobA, which pays $10 ,000 less than JobE, but has a corporate culture that I've weighted 40 % higher, I'm using a compensatory strategy.

I'm willing to trade money for satisfaction, and MAUT tells me exactly what the rational trade should be.

Exactly.

But here's the critical limitation.

While MAUT is robust as a normative model, Payne's 1976 apartment selection study show that people don't always use it descriptively, especially not when they're under cognitive strain.

Payne varied the cognitive load by manipulating the number of apartment options, 2, 6, or 12, and the number of informational factors, 4, 8, or 12.

When participants only had two apartments to choose from, they behaved rationally.

They used compensatory strategies.

They examined the same number of factors for both and were willing to make those trade -offs.

But when the number of apartments increased to 6 or 12, the sheer volume of data just led to severe cognitive strain.

And participants abruptly switched models.

They resorted to elimination by aspects, or EBA.

EBA is a descriptive model and, critically, a non -compensatory strategy.

The decision -maker selects one factor, say a rent threshold, and immediately eliminates all alternatives that exceed that limit.

Completely ignoring how well those eliminated apartments might have scored on other factors, like closet space, or noise level, or location.

So an apartment that has the perfect location, incredible amenities, a dream layout, it's instantly thrown out if its rent is $5 over your specified limit.

The good factors cannot compensate for that one bad factor.

Correct.

The decision -maker just keeps selecting factors and eliminating choices until only one option is left.

EBA is a powerful cognitive heuristic to reduce information overload, showing that when faced with complexity, people drop the rigor of MAUT for the cognitive ease of serial elimination, even if it means throwing out potentially optimal choices.

This shift takes us directly into the realm of descriptive models of decision -making models that recognize how decisions are actually made in the initial screening phases and in high -stakes, real -time contexts.

And the first of these is image theory, proposed by Beach.

This is another non -compensatory model, and it argues that most of the work happens not in utility calculations, but in the pre -choice screening of options, where we quickly reduce the alternatives to maybe one or two viable options.

And this screening is done by checking the compatibility of the option against three key images that we hold about ourselves in our future.

What are they?

First is the value image, which holds the decision -maker's core morals, principles, and values.

Second, the trajectory image, which holds their future goals, their aspirations, their desired path.

And third, the strategic image, which is all about the plans and tactics they've established to attain those trajectory goals.

And this model is non -compensatory.

If a choice is incompatible with any one of these images, it's rejected immediately.

No calculation.

Absolutely.

Think about the college major rejection examples.

Rejecting economics because you believe all econ majors only care about money is a violation of your value image.

Rejecting a career path because you feel it has no future, that's a dead -end job, is a violation of the trajectory image.

Exactly.

And rejecting a job offer because accepting it would violate your established plan to move to a specific city and attend a certain graduate school, that would be a violation of the strategic image.

These rejections happen upfront, often based on vague, intuitive criteria, making any formal MAT calculation completely irrelevant.

That's a fascinating look at the power of screening.

The other key descriptive model applies specifically to experts making lightning -fast, high -stakes decisions.

Recognition Prime Decision Making, RPD, developed by Gary Klein.

RPD is based on observing experts like firefighters, nurses in the ICU, or military commanders.

In these scenarios, the time pressure is immense, and there's zero time for a MAUT spreadsheet.

Experts rely heavily on intuition, analogy, pattern matching, and rapid mental simulation.

Can you walk us through the RPD process?

It sounds like organized intuition.

It is, in a way.

The process starts when the expert sizes up a situation,

often comparing it to past scenarios they've encountered, using intuition to identify the critical cues and features.

They typically consider only one option at a time, not six simultaneously.

They then mentally simulate the likely effect of that decision.

If the simulation fits the scenario and is expected to work, they implement it immediately.

If the mental simulation reveals a flaw, they adjust the option or find another, usually by shifting the metaphor they're using to understand the situation.

The famous firefighter anecdote from Box 13 to 6 vividly illustrates this type of rapid, expert pattern matching that RPD describes.

It's a perfect example.

In the anecdote, a lieutenant leads his crew into a house fire.

They start spraying water, but the fire roars back, which the lieutenant recognizes as odd.

Then, though he can't articulate why, he gets this sudden sixth sense, a feeling that the situation is gravely wrong and orders his men out of the building.

And moments after they exit, the floor collapses.

The floor collapsed because the fire was in the basement, burning up from underneath, which explained the odd behavior of the fire roaring back when they sprayed it on the first floor.

That sixth sense was not a mathematical computation.

It was a rapid, unarticulated pattern matching and recognition of an anomaly.

The situation failed to match the standard mental metaphor for a floor fire, and that deviation triggered an intuitive rejection.

RPD suggests that much of the competence and expert decision -making is this intuitive recognition and immediate mental simulation far removed from any algorithmic comparison.

So, we've established the landscape.

Humans are prone to these powerful heuristics that introduce systematic error, and our idealized normative models like EU and MAUT often fail descriptively because of complexity, forcing us into simpler strategies like EBA, image theory, or RPD.

This naturally leads us to the crucial final question.

How can we actually improve decision -making?

The major obstacles are clear, and they're cognitive.

First, overconfidence, which prevents people from even seeking aids in the first place, and second, the cultural preference for trusting vivid human intuition and clinical impressions over objective, numerical, statistical aids.

And the research is pretty sobering.

Simply telling people about cognitive biases,

like anchoring or framing effects and warning them not to do it, results in little to no measurable improvement.

Knowing the bias does not fix the bias.

Real improvement requires a sustained effort.

It's a combination of extensive practice,

individual feedback on performance calibration, and most importantly, clarifying the statistical or probabilistic aspects of the decisions.

Under those specific conditions, substantial reductions in bias have been reported.

A key finding in reducing bias comes from Nisbet and his colleagues back in 1983, who found that simply making the role of chance more salient encourages people to reason more statistically and less based on vivid, small samples.

They showed this beautifully with the David L.

College visit example from Box 13 -7.

In the first version of the story, David L.

visits an Ivy League school, has a great time, and decides to attend, ignoring his many friends' prior warnings about how unhappy they were there.

And participants in the study tended to side with David L.'s vivid, positive experience.

But in version 2, they added one critical detail.

They explained that David L.

had randomly selected the classes, activities, and people he visited.

And by describing David L.'s personal experience as a randomly selected sample,

participants gave significantly less weight to the vivid, small sample of his campus visit.

They relied more rationally on the larger, statistically sound sample of his friends'

consistent negative advice.

Simply making the role of chance explicit helped people activate their statistical reasoning and disregard the compelling yet unreliable anecdote.

This evidence brings us back to that inherent tension we see between clinical intuition and formal statistical models.

Mill famously examined the relative effectiveness of holistic clinical impressions versus judgments made by statistical formulas.

Right.

Think of a hiring manager predicting an employee's success versus a formula weighing standardized test scores, GPA, and recommendation strength.

The findings consistently, overwhelmingly support the use of non -human statistical models in predictions.

Whether it's college success, medical diagnoses, or criminal recidivism.

This challenges our core perception of authority.

Why does a cold algorithm beat a warm, experienced human expert?

Dawes provided the essential insight here.

Humans are actually quite proficient at figuring out which variables are the good predictors, the content selection.

We know GPA and SAT matter.

The human shortcoming appears when we try to integrate all that information in an unbiased and consistent way.

Our subjective mood, our biases, our overconfidence, they all contaminate the weighting and integration process.

Exactly.

Linear models like those used in M .A .K .E.

analysis are excellent at unbiased, consistent integration.

They don't have bad days or get tired.

They apply the weights exactly as specified every single time.

Therefore, the emerging solution, the one that bridges the gap between human insight and mathematical consistency, is decision analysis.

Decision analysis, as described by Keeney, is an emerging technology that combines the best of both worlds.

It relies on human judgment to select the relevant variables,

determine the subjective beliefs, the utilities and probabilities, and assign the personal weights.

But then it uses unbiased, consistent integration methods, the algorithm, to structure the decision and arrive at the most rational choice.

It allows people to approach complex decisions with the care and structure they deserve, mitigating our human tendency toward bias and inconsistency.

That's such a comprehensive view of the entire landscape.

We began with the five phases of decision making, detailed the powerful yet deeply flawed nature of cognitive heuristics that emerged from cognitive overload, navigated the distinction between our idealized normative models and our actual descriptive behavior, and concluded with the clear evidence that structured aids like decision analysis lead to demonstrably more rational outcomes.

And so here is a final provocative thought for you to chew on.

Given the overwhelming evidence that our human intuition is prone to confidence,

systematic biases, and the avoidance of necessary trade -offs, when you're faced with your next major life decision, the one involving high stakes and many variables, will you trust your gut?

Or will you take the time to formally structure the decision and use objective weights, even if that process feels less human and takes uncomfortable time away from the immediate choice?

Think about the power of that choice.

Thank you for joining us for this deep dive into the cognitive process of decision making.