1.1 System 1 and System 2
This model is concerned with how people make decisions and choices.
What are the ingredients of a good/ rational decision? What are the steps we should go
through? How Econs decide…
1. Define the problem
2. Identify all decision criteria
3. Allocate weights to the criteria
4. Identify all alternatives
5. Evaluate the alternatives
6. Choose the best alternative
Even though this might be the rational way to go about making a decision, people don’t
usually go through all these steps. Instead of identifying all alternatives, people might only
visit one store instead of all of them. The agents that go through these steps when making
decisions are called Econs, because this tends to be the way the science of economics
conceptualized decision making -> there are rational agents and when they make a decision,
they go through all these steps.
Beyond rational choice
However in this course we are more interested in how humans decide. Let’s move beyond
rational choice, and look into more realistic decision making processes.
- Bounded rationality
Resources (time, cognitive, etc.) are limited. We often don’t optimize, but satisfice (we
consider only a few alternatives). Because of bounded rationality we fail to make the best/
rational choice.
- Predictable irrationality
We are not just often wrong. We are systematically wrong (these aren’t random patterns ->
we can predict the way in which people are wrong in their choices). This is important,
because being able to predict the ways in which people are wrong, means we can study how
we decide and possibly find ways to improve decision making.
-> We tend to make choices which are biased, which means they are systematically wrong in
some precise directions.
System 1 and system 2
Example: A bat and a ball cost $1.10 in total. The bat costs $1 more than the ball. How much
does the ball cost? … cents
You might be inclined to think it is 10 cents. This is however wrong, it would make the price
$1.20. You automatically first go to 10 cents and then you have to correct yourself as you
realize 10 is not the correct answer. This process bringing you to the 10 cents is system 1.
System 1 (intuitive answer) System 2 (rational answer)
- Automatic - Controlled
- Fast - Slow
- Not cognitively demanding - Cognitively demanding
Example: 10 cents Example: by using the rules of addition, you
would come to the right answer: 5 cents.
System 2 makes you pay more attention to the
problem you are facing.
,System 1 and system 2 thinking is a way for us to understand decision making processes. It
is a simplification/ a model to help us understand how people make decisions. Through
system 1 we get an intuitive answer and system 2 then allows us to question the intuitive
answer we obtained from system 1 (allows us to ask ourself: is it the correct answer?). If
system 1 does not give us the right answer, system 2 allows us to get to the rational answer
by applying rules (if we are resourceful enough).
Heuristics and biases
We can apply system 1 and system 2 thinking to heuristics and biases.
- System 1 = heuristics
System 1 most of the time is responsible for the use of heuristics; these are mental
shortcuts to satisfactory solutions. Rather than calculating what the optimal solution/
response is to a problem, system 1 uses mental shortcuts which are usually correct.
Why are these heuristics useful?
System 1 is very efficient, no resources are required, they are intuitive answers. It also tends
to be the case that they work most of the times. Heuristics pays off because it allows us to
reach satisfactory solutions while not using much cognitive resources.
Heuristics work most of the times, so sometimes they don’t. And when they don’t work they
lead to biases. This goes back to the systematically, predictably wrong answers. Biases are
essentially systematic deviations from rationality/ optimality. So a bias is a systematic,
predictable mistake. Back to the example: people who don’t correctly identify 5 cents as the
answer, they don’t just come up with a wrong answer, but they tend to say 10 cents -> so we
can predict they will say 10 cents. This is the result of a bias. As a result of using a mental
shortcut (that might work most of the times) we might end up with the wrong answer and that
wrong answer is predictable/ systematic and therefore we can study it.
Two notes of caution:
1. ‘What is rational’ is a debated question.
2. Not all ‘biases’ are due to heuristics/ system 1, some might be system 2.
---------------------------------------------------------------------------------------------------------------------------
,1.2 Heuristics in Probability Estimations: Representativeness
We make probability estimates all the time. We estimate probability for important decisions,
e.g. buying a house, vacation destination, whether we are sick, how much we will like a
product etc. All these judgements involve estimating a probability. The question then
becomes:
How do we make these kinds of judgements?
What are some of the heuristics we use?
Representativeness heuristic
“The more X resembles Y, the more likely X is to be Y”.
“When I see a bird that walks like a duck and swims like a duck and quacks like a duck, I call
that bird a duck”.
This sounds right and isn’t inherently wrong, so we can agree with this statement and most
of the times it would tend to be true.
However the problem with this statement is that it tends to ignore some information that
might be relevant. That way you might fail to ask yourself the following 2 questions:
- How confident are you in this resemblance?
● Accurate stereotypes are still stereotypes
Are you sure there is no stereotype?
E.g. CEO riddle exposes a stereotype (CEO = man). Heuristics exist because
they tend to work just fine in many situations -> the connection between a
stereotype and a heuristic tends to be true: 95% CEOs are male, only 5% are
women. People failed to solve the riddle correctly because they had a
stereotype in mind. Stereotypes are misleading and make people dumber ->
people couldn’t solve an easy riddle.
This stereotype is an oversimplification (CEO=male), if you base yourself on
this judgment, you are going to be wrong in estimating the probability.
● Information is often not enough, we need more information
Through stereotyping you are ignoring relevant information. We need a lot of
information to calculate resemblance accurately.
Example:
The smaller a sample is, the less likely your observations are representative
of the entire population.
Many are wrong and thought the probability would be about the same/ equally
likely chance to have more boys as we are generally insensitive to sample
sizes.
, -> We underestimate the importance of information -> we need more
information. If you try to estimate the probability that X belongs to Y based on
how resembling X and Y are -> are we then really confident in this
resemblance? So basing yourself on resemblance, is inherently limiting, to
the extent that you don’t have enough information about the resemblance.
- How likely is Y in the first place?
Many times we forget that the likelihood that X is Y is also dependent on the probability of Y
itself. If you base yourself on representativeness, you are ignoring this relevant probability.
● Representativeness ignores base rates
If you evaluate the probability that X is Y based on how much X resembles Y,
you are not asking yourself how likely is Y in the first place.
Example:
1 is the right answer. 2 is a conjunction of 2 independent categories of people:
bank tellers and feminists -> feminist bank tellers are the conjunction of 2
events. Blue and orange on its own are larger than the overlapping part -> so
by definition the probability for any individual to be a feminist bank teller is
smaller dan being a bank teller. Linda resembles 2 very much, but by
definition 2 is less likely than 1 -> this is the conjunction fallacy. The
resemblance blind people for the fact that 2 is less likely than 1. By definition
the base rate is different and therefore it must be more likely that Linda is only
a bank teller as opposed to her being a feminist bank teller, just because 2 is
a subset of 1 -> meaning the base rate is by definition a subset of the first
one.
● False positives
Example:
The fact that only 10% of people are misclassified (false positive), makes it
highly unlikely in our mind that we get misclassified, while in fact it is quite
likely even when tested positive we actually do not have antibodies.
---------------------------------------------------------------------------------------------------------------------------
This model is concerned with how people make decisions and choices.
What are the ingredients of a good/ rational decision? What are the steps we should go
through? How Econs decide…
1. Define the problem
2. Identify all decision criteria
3. Allocate weights to the criteria
4. Identify all alternatives
5. Evaluate the alternatives
6. Choose the best alternative
Even though this might be the rational way to go about making a decision, people don’t
usually go through all these steps. Instead of identifying all alternatives, people might only
visit one store instead of all of them. The agents that go through these steps when making
decisions are called Econs, because this tends to be the way the science of economics
conceptualized decision making -> there are rational agents and when they make a decision,
they go through all these steps.
Beyond rational choice
However in this course we are more interested in how humans decide. Let’s move beyond
rational choice, and look into more realistic decision making processes.
- Bounded rationality
Resources (time, cognitive, etc.) are limited. We often don’t optimize, but satisfice (we
consider only a few alternatives). Because of bounded rationality we fail to make the best/
rational choice.
- Predictable irrationality
We are not just often wrong. We are systematically wrong (these aren’t random patterns ->
we can predict the way in which people are wrong in their choices). This is important,
because being able to predict the ways in which people are wrong, means we can study how
we decide and possibly find ways to improve decision making.
-> We tend to make choices which are biased, which means they are systematically wrong in
some precise directions.
System 1 and system 2
Example: A bat and a ball cost $1.10 in total. The bat costs $1 more than the ball. How much
does the ball cost? … cents
You might be inclined to think it is 10 cents. This is however wrong, it would make the price
$1.20. You automatically first go to 10 cents and then you have to correct yourself as you
realize 10 is not the correct answer. This process bringing you to the 10 cents is system 1.
System 1 (intuitive answer) System 2 (rational answer)
- Automatic - Controlled
- Fast - Slow
- Not cognitively demanding - Cognitively demanding
Example: 10 cents Example: by using the rules of addition, you
would come to the right answer: 5 cents.
System 2 makes you pay more attention to the
problem you are facing.
,System 1 and system 2 thinking is a way for us to understand decision making processes. It
is a simplification/ a model to help us understand how people make decisions. Through
system 1 we get an intuitive answer and system 2 then allows us to question the intuitive
answer we obtained from system 1 (allows us to ask ourself: is it the correct answer?). If
system 1 does not give us the right answer, system 2 allows us to get to the rational answer
by applying rules (if we are resourceful enough).
Heuristics and biases
We can apply system 1 and system 2 thinking to heuristics and biases.
- System 1 = heuristics
System 1 most of the time is responsible for the use of heuristics; these are mental
shortcuts to satisfactory solutions. Rather than calculating what the optimal solution/
response is to a problem, system 1 uses mental shortcuts which are usually correct.
Why are these heuristics useful?
System 1 is very efficient, no resources are required, they are intuitive answers. It also tends
to be the case that they work most of the times. Heuristics pays off because it allows us to
reach satisfactory solutions while not using much cognitive resources.
Heuristics work most of the times, so sometimes they don’t. And when they don’t work they
lead to biases. This goes back to the systematically, predictably wrong answers. Biases are
essentially systematic deviations from rationality/ optimality. So a bias is a systematic,
predictable mistake. Back to the example: people who don’t correctly identify 5 cents as the
answer, they don’t just come up with a wrong answer, but they tend to say 10 cents -> so we
can predict they will say 10 cents. This is the result of a bias. As a result of using a mental
shortcut (that might work most of the times) we might end up with the wrong answer and that
wrong answer is predictable/ systematic and therefore we can study it.
Two notes of caution:
1. ‘What is rational’ is a debated question.
2. Not all ‘biases’ are due to heuristics/ system 1, some might be system 2.
---------------------------------------------------------------------------------------------------------------------------
,1.2 Heuristics in Probability Estimations: Representativeness
We make probability estimates all the time. We estimate probability for important decisions,
e.g. buying a house, vacation destination, whether we are sick, how much we will like a
product etc. All these judgements involve estimating a probability. The question then
becomes:
How do we make these kinds of judgements?
What are some of the heuristics we use?
Representativeness heuristic
“The more X resembles Y, the more likely X is to be Y”.
“When I see a bird that walks like a duck and swims like a duck and quacks like a duck, I call
that bird a duck”.
This sounds right and isn’t inherently wrong, so we can agree with this statement and most
of the times it would tend to be true.
However the problem with this statement is that it tends to ignore some information that
might be relevant. That way you might fail to ask yourself the following 2 questions:
- How confident are you in this resemblance?
● Accurate stereotypes are still stereotypes
Are you sure there is no stereotype?
E.g. CEO riddle exposes a stereotype (CEO = man). Heuristics exist because
they tend to work just fine in many situations -> the connection between a
stereotype and a heuristic tends to be true: 95% CEOs are male, only 5% are
women. People failed to solve the riddle correctly because they had a
stereotype in mind. Stereotypes are misleading and make people dumber ->
people couldn’t solve an easy riddle.
This stereotype is an oversimplification (CEO=male), if you base yourself on
this judgment, you are going to be wrong in estimating the probability.
● Information is often not enough, we need more information
Through stereotyping you are ignoring relevant information. We need a lot of
information to calculate resemblance accurately.
Example:
The smaller a sample is, the less likely your observations are representative
of the entire population.
Many are wrong and thought the probability would be about the same/ equally
likely chance to have more boys as we are generally insensitive to sample
sizes.
, -> We underestimate the importance of information -> we need more
information. If you try to estimate the probability that X belongs to Y based on
how resembling X and Y are -> are we then really confident in this
resemblance? So basing yourself on resemblance, is inherently limiting, to
the extent that you don’t have enough information about the resemblance.
- How likely is Y in the first place?
Many times we forget that the likelihood that X is Y is also dependent on the probability of Y
itself. If you base yourself on representativeness, you are ignoring this relevant probability.
● Representativeness ignores base rates
If you evaluate the probability that X is Y based on how much X resembles Y,
you are not asking yourself how likely is Y in the first place.
Example:
1 is the right answer. 2 is a conjunction of 2 independent categories of people:
bank tellers and feminists -> feminist bank tellers are the conjunction of 2
events. Blue and orange on its own are larger than the overlapping part -> so
by definition the probability for any individual to be a feminist bank teller is
smaller dan being a bank teller. Linda resembles 2 very much, but by
definition 2 is less likely than 1 -> this is the conjunction fallacy. The
resemblance blind people for the fact that 2 is less likely than 1. By definition
the base rate is different and therefore it must be more likely that Linda is only
a bank teller as opposed to her being a feminist bank teller, just because 2 is
a subset of 1 -> meaning the base rate is by definition a subset of the first
one.
● False positives
Example:
The fact that only 10% of people are misclassified (false positive), makes it
highly unlikely in our mind that we get misclassified, while in fact it is quite
likely even when tested positive we actually do not have antibodies.
---------------------------------------------------------------------------------------------------------------------------
