bias and decision making Flashcards
1
Q
define rationality
A
when decision making appears to be rational
2
Q
norms & rationality
A
- norms are rules of action or thought which define optimality
- rationality is a set of norms
–> be consistent (‘coherence’)
–> correspond to reality (‘correspondence’) - but nobody can agree on the complete set of norms for reasoning
3
Q
the norms of rationality
A
- coherence
–> be consistent - correspondence
–> correspond to reality
4
Q
two types of bias
A
- availability bias
- framing bias
5
Q
availability bias
A
- over estimating the frequency of rare of memorable events
–> e.g. overestimating the frequency of plane crashes
6
Q
framing bias
A
- switching decision based on the question framing
–> e.g. would you buy a new ticket if you dropped it on the way to the cinema?
vs
–> if you dropped £10 on the way to the cinema would you buy another ticket?
7
Q
violation of norm and conjunction fallacy
A
- conjunction fallacy = have made an error in reasoning by assuming that specific conditions are more probable than a single, more general condition
–> e.g. Linda problem
8
Q
the Linda problem and what it shows
A
- Linda is 31 years old, single, outspoken, and very bright
- She studied philosophy at University
- As a student, she was deeply concerned with issues of discrimination and social justice, and also participated in anti-war demonstrations
- Which is more likely?
1. Linda is a bank teller (a person working in a bank)
2. Linda is a bank teller and is active in the feminist movement - most people pick 2 despite norms (especially probability)
- option 2 has to be less likely than option 1 because it is a subset of option 1
–> option 2 require two things to be true, option 1 only requires one
9
Q
how should we make rational decisions? (decision calculus)
A
- be logical
- use probabilities
- systematic consideration of all options
10
Q
value and utility (probability and decision making)
A
- difficulty in making decisions:
1. future is uncertain
2. need to assess potential risks and benefits
3. choose a course of action to increase the chance of a positive outcome
4. need to use knowledge to estimate probabilities of future events
11
Q
probability based on value
A
- expected value = value an investment will have in the future
- good bet = expected value is greater than amount invested
- bad bet = expected value is less than amount invested
- rational choice = invest to maximise expected value
- value is not always the most important factor in our decision making
12
Q
what is risk aversion?
A
tendency for people to accept a sure outcome over a riskier outcome
13
Q
expected utility theory
A
- calculating the option with the highest expected utility is a decision-making method
- rational decision making may assume that we choose the option that maximises our utility
14
Q
premise of the expected utility theory
A
- expected utility theory is a theory of decision under risk
- each option leads to one set of outcomes
- where the probability is known
15
Q
value is not utility
A
- utility is compressed with respect to value
–> you don’t enjoy 10 pizzas 10x more than 1 pizza (although 10 pizzas costs 10x more) - value correlates with cost more than utility
- utility is how much you enjoy / prefer it = the degree to which it contributes to well being and satisfies your desires
- utility can be measured as ‘willingness to pay’
- according to expected utility theory people should behave to maximise expected utility
–> even if value is lower
16
Q
marginal utility
A
- diminishing marginal utility of wealth:
–> basic idea = as the amount of money one has increases, each addition to one’s fortune becomes less important, from a personal, subjective point of view
–> extra £1000 means very little to Bill Gates but an extra £1000 to a uni student means quite a lot
17
Q
uncertainty affects expectation
A
- pizza example:
–> If you would pay £10 for a pizza, how much would you pay for…
–> a 10% chance of winning a pizza?
–> a 1/100 chance that the pizza made you sick? - we assign our expected utility to these
–> we go with the one with the greater expected utility
–> pay more for 10% chance than the 1/100
–> it costs more but has greater utility
18
Q
equation for calculating expected utility
A
- E = P x U
–> E = expected utility
–> P = probability
–> U = utility
19
Q
example of expected utility
A
- expected utility of a bet with a 50% chance of a £40 pay off?
- answer = £20
–> ( E = 0.5 x 40)
20
Q
calculating expected utility with multiple options
A
- E = (P1 x U1) + (P2 x U2)
- EXAMPLE:
–> lottery offered for £15
–> 20% chance of winning £50
–> 10% chance of winning £100
–> do you buy?
–> (expected utility has to exceed the cost) - ANSWER = YES
–> P1 x U1 = 10
–> P2 x U2 = 10
–> answer = 20 and 20 exceeds 15
21
Q
the effect of uncertainty
A
- everything is uncertain question is just how much
- can have scenarios where chances of high and low utility is equal
–> two separate and equal chunks near high and low - can have scenarios where things are probably good but there is a chance they can go bad
–> larger chunk of graph near high - can have scenarios where things are likely to be okay but could go VERY good or VERY bad
–> one chunk spread from high to low, peak is mid
22
Q
Bias: loss aversion and prospect theory
A
- often when there is guaranteed loss, people which choose the option where they avoid this over expected utility
–> loss aversion
–> people pick options where loss may or may not happen, over option where loss is guaranteed, regardless of expected utility
–> having £100 and losing £50 is worse than starting with £0 and gaining £50 - we have an innate motive to avoid loss rather than achieving similar gain
- we feel pain of loss more than pleasure of gain
- thus we have a bias in decision making
23
Q
Do we use norms, probabilities and EU in real world decision making?
A
- Do we always have the knowledge and time to use such methods of decision making in an uncertain world?
–> links to heuristics
24
Q
rationality is bounded
A
- world is complex
- decisions need to be made quickly
- our time is limited
- our cognitive capacities are limited
25
Q
ecological rationality
A
- evolutionary perspective as to why we use heuristics
- rationality is defined by how you should behave in the environment to survive rather than simply by the norms
- correspondence (with the environment - how cognition works in the world) is more important than coherence (logical processing)
–> behaviours from the past may fail when the environment changes rapidly
26
Q
adaptive value
A
- the value of an action across evolutionary time
- evolution maximises long term expected value
- errors are permissible if average benefit
- avoiding costly mistakes
- avoid high costs (including decision costs)
–> example = we still look when crossing the road even if road is clear and quiet
27
Q
what do heuristics make assumptions about?
A
- the environment
- what things go together (association)
- probabilities (risk)
28
Q
what are heuristics? function?
A
- general rule of thumb
–> based on instinct, emotions, environment etc… - give us strategies in an uncertain world
- they make assumptions about the world
–> i.e. what keeps us safe - allows us to make decisions when we are unable to use systematic strategies
- they do not attempt to find the optimal solution but a good enough solution
- but they can have biases
29
Q
recognition heuristic?
A
- this works when you have some (but not no) knowledge
- works when the likelihood of hearing an option correlates with its value/importance/size
- it doesn’t work when you know too much or too little
- when incidence isn’t related to value it can lead to systematic errors
- successful heuristics need to be ‘fast and frugal’
–> frugal as they ignore part of the information
30
Q
are all biases flaws?
A
- no
30
Q
biases / errors defined by norms
A
- expected utility theory
- laws of logic
- laws of probability (e.g. conjunction problem)
–> deviation from the above norms/laws can be seen as error - OR they can be seen as evolutionary adaptiveness (ecological rationality)
–> strategies have evolved and are adaptive so over evolutionary time they become more successful than not
–> so there are systematic errors in decision making but the errors are more beneficial than not
–> hence biases shouldn’t be seen as irrational or as flaws
31
Q
Wason’s 2-4-6 task
A
- task: find the generating rule I am using to create 3 number sequences. You can propose test examples and be told if they fit or break the rule
- starting example: 2-4-6
- what example will you try next?
32
Q
Wason’s 2-4-6 task example
A
- true generating rule: - successive even numbers
- first test: 2 - 4 - 6
–> answer: “fits the rule” - second test: 4 - 6 - 8
–> answer: “fits the rule” - third test: 4 - 6 - 9.
–> answer: “doesn’t fit!” - guess: “It’s successive even numbers”
33
Q
Wason’s 2-4-6 task general results
A
- result: most people, most of the time, think of a highly specific rule (e.g. ascending even numbers) and generate an example which fits (e.g. 10-12-14)
–> this a confirmation bias
–> a ‘positive test strategy’
34
Q
types of test strategies
A
- positive tests:
–> seek to verify the hypothesis (but can find it false)
–> e.g. saying ‘6-8-10’ if you think the rule is ascending even numbers - negative tests:
–> seek to falsify hypothesis (but can find it true)
–> .g. saying ‘3-5-7’ if you think the rule is ascending even numbers
35
Q
positive vs negative strategies, which is best?
A
- depends on what you believe and how it relates to the truth
- general rules:
–> positive strategies will always confirm your prior belief, belief never improves
–> not the best
–> if you think it is specific, but it’s general, you keep saying specific things and won’t change your belief - specific rules:
–> positive strategy rapidly converges on the truth
–> if you think it’s specific and it is specific, you will prove it
36
Q
Wason task and error in understanding
A
- 2-4-6 task was the first experiment that showed people to be illogical and irrational
37
Q
everyday example of different test strategies
A
- knock at door
- hypothesis: someone is at the door
- positive: open the door
- negative: check every other source of knocking
- positive is the best in this case
- fit between hypothesis and reality determines which is best
38
Q
alternative explanations for errors
A
- can errors in decision making tasks be accounted for in a way that does not suggest thinking is not logical?
- communication interpretation:
–> participants assume that all of the information is provided for a reason so they use that information to make their decision
–> like in the Linda task
–> participants may not be seeing the case as a problem solving situation
–> thus they want to use all of the information
–> if it is there, is it really an error?
39
Q
biases are not mere errors
A
- mistakes are one-off
- systematic errors are consistent
- biases are making the wrong choice but for a reason
- just like visual illusions reveal the normal mechanisms of perception heuristic errors reveal the normal mechanisms of reasoning
40
Q
reasons we might mistake systematic errors
A
- using a strategy optimised for a different environment
- considering a (different) bundle of choices
- using a different cost/benefit analysis
41
Q
heuristic vs rational thinking
A
- heuristic / system 1:
–> quick
–> automatic
–> effortless
–> unconscious - rational / system 2:
–> slow
–> deliberate
–> effortful
–> conscious
42
Q
rational vs heuristic thinking - dual process
A
- system1 and system 2 is a dual process theory
- there are many of them
- predictions:
–> features should go together (so, for example, no conscious heuristic thinking, no unconscious rational thinking)
–> people should switch between modes of thinking in the right circumstances (more time = more deliberative thinking)
43
Q
dual process theories
A
- a psychological theory which posits two processes which work at the same time, or support the same capacity:
–> usually work together
–> can be in conflict with each other
–> one usually has system 1-like properties, one system 2-like properties
44
Q
where you will see biases
A
- decision biases we discussed today are features of normal cognitive processes
–> e.g. memory, attention, planning - can affect experts and laypeople alike
- biases are part of thinking, not an optional extra
45
Q
implications for investigating bias
A
- If you see a bias try and understand how people understand a decision:
–> their perception of costs & benefits
–> their experience of similar decisions
–> their assumptions about the environment - do not conclude they are irrational
46
Q
key points:
A
-Biases are defined against norms (standards of rationality)
- Expected Utility Theory (E=p*U) defines one standard, and makes predictions about how uncertainty should affect decisions
- Heuristics are designed to be ecologically rational - to provide quick, good enough choices
- Biases are not flaws in reasoning, they are a way of understanding how reasoning is done
- Dual process theories suggest we have a fast system 1 and a slow system 2
