1
Q
what is rationality?
A
- ultimate function: evolution by natural selection shaped our minds to solve survival-relevant problems
- proximate mechanism: computation implemented in our brains provides the means by which we solve those problems
- rationality links these 2: explains why the computations are adaptive
2
Q
what is the difference between epistemic rationality and instrumental rationality?
A
- epistemic: reasoning processes that reliably lead to true beliefs about the world
- instrumental: decision-making processes that lead one’s goals being achieved
3
Q
how does epistemic/instrumental rationality come apart?
there are 3 effects
A
- optimistic beliefs: creates self-fulfilling prophecies
- placebo effect: false beliefs about a treatment can promot health
- expressing false beliefs: solidify social ties if shared with one’s group
4
Q
why do we depart from ideals?
there are four possibilities
A
- bounded rationality
- ecological rationality
- evolutionary misadaptation
- the formal models are wrong
5
Q
what is bounded rationality?
A
we face a range of cognitive limits:
- working memory only has ~7 chunks of info
- can only pay attention to ~1 thing at a time
- cannot calculate instantaneously –> sometimes we can’t wait
6
Q
according to the bounded rationality, what does it mean that we are as rational as we can be?
A
- we are rational, relative to those limits
- if you are choosing among 10 different kinds of toothpaste, each with 5 different characteristics, you would need a spreadsheet to “optimize” this decision because we couldn’t keep track of so much infomration
- instead we satisfice – we pick an option that is good enough
7
Q
what is ecological rationality?
A
- we have evolved adaptive solutions to many problems we face in the real world, but depart from ‘textbook’ problems or lab experiments
- ‘irrational’ biases in unusual contexts can result from cognitive mechanisms that work well in everyday settings for which they are adapted
8
Q
what is an example of the ecological rationality?
when catching a ball
A
catching by calculation vs by gaze
- calculation is very complicated
- the gaze method: fixate your eyes on the ball, run, and adjust your speed so that the angle of gaze remains constant (many predators use this method but it only works under the right conditions - doesn’t work underwater)
9
Q
what is evolutionary misadaptation?
A
- maybe we’re adapted well to solve the problems our ancestors faced in the environment of evolutionary adaptedness (EEA)
- problems in perception and langauge were very similar to EEA to those we face today, but others are not
calories are abundant, environment is less threatening, tech and commerical culture change the decision landscape
10
Q
what does “the formal models were wrong” mean in rationality?
A
- risk: we can calculate probabilities because we can list all possible outcomes and we know the model that generates them
- radical uncertainty: we can’t calculate probabilities; we simply do not know
11
Q
what is intuition?
A
- a quick response that effortlessly comes to mind
- often these are right - but sometimes they mislead us
12
Q
what is metacognition?
A
“thinking about thinking” - monitoring and controlling thoughts and feeling. corrects any mistakes that our intuition makes
13
Q
what is dual process theory?
A
- system 1: some processes are quick, intuitive, automatic
- system 2: others are slow, reflective, effortful
- system 2 monitoring system 1: metacognition if an effortful process
14
Q
what is the cognitive reflectiveness test?
A
- measures differences in motivation to question intuitions
- ability for system 2 (metacognition) to monitor and override system 1 response (intuition)
higher CRT score is associated with: more risk-neutral prefs and more willingness to delay rewards
15
Q
what is pseudoprofound bullshit?
A
- how peolpe respond to statements like “wholeness quiets infinite phenomena”
- people who found this profound have lower: cognitive reflectiveness, verbal intelligence, numeracy, susceptibility to various cognitive biases
these results didn’t hold up for actually profound statements: a wet person does not fear the rain
16
Q
what is heuristics?
A
- a menatl shortcut that solves a problem without carrying out a full analysis
- heuristics save effort
- they might or might not give similar answers to full analysis
- a key part of “bounded rationality” - finding the best solution subject to limits of time, memory, and information
17
Q
what is bias?
A
- a systematic error caused by applying a heuristic
- major debate in this field is whether heuristics are consistently error-prone
some people think that heuristics inevitably lead to mistakes
17
Q
what is the difference between a heuristics attribute and a target attribute?
A
- target attribute: low in both evaluability (hard to estimate) and accessibility (doesn’t come to mind)
- heuristic attribute: easy to evaluate and access. includes natural assessments (e.g., size, similarity, mood, fluency)
18
Q
what is the availability heuristic?
A
a mental shortcut where people estimate the likelihood of an event by how easily examples come to mind
19
Q
what is representativeness heuristic?
A
a mental shortcut where people estimate the probability of an event based on how closely it resembles a mental prototype or stereotype of a situation
20
Q
what is fluency heuristic?
A
a mental shortcut where people judge the value of something based on how easily or quickly they can process the information
21
Q
what is social heuristics?
A
mental shortcuts or rules of thumb that people use to make quick, efficient decisions in social situations, especially when they lack time or information
22
Q
is a fully analysis necessarily better?
A
- sometimes simpler solutions actually work better - they’re not only simpler but can also be more accurate
- if you’ve ever taken a stat or machine learning course, you might have come across the idea of overfitting: if you use too many pedicators in a model, it states to fit not just the true phenomena, but also the random noise
23
Q
what is multiple regression?
A
a statistical technique that attempts to calculate the optimal weights for every cue
24
what is "take the best"?
bases a prediction on just the single most valid cue
25
what is "tallying"?
just count up the number of positive and negative cues
26
what is the level of accuracy between multiple regression, "take the best", and tallying?
take the best > tallying > regression
27
what is recognition heuristic?
a mental shortcut where people make inferences by choosing the option they recognize over one they do not, assuming the recognized option is more likely to be better
## Footnote
people from foreign countries are often better than citizens of a country at prediciting populations of cities
28
what is the analytical reason as to why people prefer to eat candy over vomit?
stimulus (bacteria) --> perception (characteristic vomit smell) --> emotion (disgust) --> behaviour (avoidance)
29
what are the 3 functions of emotion?
1. help make snap judgements: often faster than conscious processing + informationally encapsulated (hard to control top-down) + sometimes take the form of "prepared" reactions (innate dislike of vomit)
2. assessing and pursuing value: cognition alone doesn't tell us what to do (what hume said) - we need to integrate our bliefs with our needs and goals. we saw that even nematodes need some basic affective states (valence) to implement the simplest steering algorithm
3. regulating social relations: we wear our emotions on our face - often involuntarily. communicating to others about threats in the environment and communicating truthfully about one's mental states to others
30
what is the somatic marker hypothesis?
- physiological origins of emotion
- feelings experienced in the body and interpreted by the brain guide decision-making
31
what is the appraisal theory?
- cognitive origins of emotions
- emotions result from cognitive appraisals of a situation, which lead to characteristic action tendencies (fear --> flee ; anger --> fight ; interest --> gather info)
32
in what way is emotion irrational?
in that they often bypass conscious thought, but can be rational in the deeper sense of carrying out the programmed agenda of natural selection
33
what do our visual systems say about rationality in lower cognition?
- our visual systems are doing the same thing all the time - inferring a 3D world froma 2D array of light on our retinas
- this would be impossible without making unconscious - and fallible - assumptions
34
what does shading gradients tell us?
the visual system assumes that light comes from above
35
what do depth cues tell us?
- it is a course of information about depth, depending on unconscious assumptions about the structure of the world and our relation to it
- these cues can fail when the assumptions are false - but it's okay because the assumptions are usually true (ecological rationality)
36
what is the different between the 3 different types of depth cues?
| (oculomotor, monocular, binocular)
- oculomotor: based on the position of our eyes and tension in our eye muscles
- monocular: rely only on the visual information available from one eye
- pictorial: work from a static image
- dynamic: depend on movement (of the viewer or the object)
- binocular: rely on visual information from both eyes
37
what are the two types of oculomotor cues?
- convergence: the inward movement of the eyes when looking at a very close object
- accommodation: how lens changes shape when focusing on near vs distant objects
38
what are the 4 different pictorial cues?
- occlusion: objects that are partially hidden are farther
- relative height: objects higher in the field of view are farther
- relative size: smaller obejcts of the same type are farther
- perspective convergence: parallel lines converge with distance
39
what are the 2 different dynamic cues?
- motion parallax: when we move, nearer objects appear to move more rapidly
- deletion and accretion: when we move, farther objects are partially covered and uncovered by nearer objects
40
what is a binocular cue?
binocular disparity: your two eyes see slightly different images and views of closer objects differ more
41
how do optical illusions work?
- depth cues usually converge on the same solution (because they are based on assumptions that are usually true)
- but they can come apart in unusual situations like optical illusions
- ecologically rational cues fail when we are taken out of the environment for which our problem-solving was adapted
42
what does memory tell us about rationality in lower cognition?
- we are likelier to forget informtion after a longer delay since practicing it
- if we forgot something, it is faster to relearn it if we practiced it more time
- our memory may prioritize information that is likely to be needed again
- information that has been retrieved more often is more likely to be needed
- information that has been retrieved more recently is more likely to be needed
43
what is deductive logic?
- a system of inference that guarantees the consistency of beliefs
- from a set of premises, we can deduce conclusions that follow logically
- if premises are true, the conclusions are guaranteed to also be true
- it doesn't tell us whether the initial premises are true
44
what makes an argument valid/sound?
- an argument is valid if the conclusion follows from the premise: the premises need not be true for an argument to be valid
- an argument is sound if it is valid AND the premises are true: the conclusions of sound arguments must be true
44
what is a proposition?
a claim about the world that can be true or false
45
what are logical connectives?
used to build complex proposiitons from simple propositions. there are 5: &, NOT, OR, IF...THEN, IF AND ONLY IF
## Footnote
truth of complex propositions can be fully determined from the truth of the simple propositions that compose them
46
what is a truth table and how does it work?
a method for determining the truth conditions of complex propositions
1. enumerate all "possible worlds" - every combination of truth and falsity for each simple proposition
2. apply the definitions of the logical connectives to determine the truth of the more complex propositions
47
what are inference rules?
rules that can be used to draw conclusions form the premises (i.e., guaranteeing the truth of conclusions if the premises are true)
48
what are valid inference rules?
| there are 3
- modus ponens: if we have P and P->Q, also gives us Q
- modus tollens: if we have -Q and P->Q, then we have -P
- de morgan's law: if we have -(P&Q), then we have -P | -Q
49
what are invalid inference rules?
- affirming the consequent: Q and P->Q gives us P
- denying the antecedent: -P and P->Q gives us -Q
50
what is the role of truth tables?
- we can use truth tables to check whether an inference rule in valid or not
- consult all rows of the truth table where the premises are true - if it's valid inference rules, the conclusion is also true
51
why are people able to reason logically?
- it's adaptive - true beliefs are useful, and deduction expands the set of true beliefs we have access to
- seems to underlie abilities we know we have, such as language
- could be the basis of domain-general thinking - being able to think abtractly about different kinds of content
- useful for manipulating concepts
52
what are concepts and propositons?
- concepts: SNOW, WHITE, REFLECTION, LIGHT
- propositions: "snow is white," "white thigns reflect light"
53
what are the two ways that we carry out deductvie reasoning?
- mental rules theory
- mental model theory
54
what is the mental rules theory?
people solve deduction tasks by:
1. converting the premises to their logical form;
2. applying a set of inference rules to those premises to construct mental proofs;
3. translating the conclusion of the proof back into the content of the premises
## Footnote
requires inference rules, a procedure for applying the rules, and a procedure for translating to/from logic and content. bounded rationality theory of logic: we follow logical rules but make mistakes due to cognitive limits like working memory
55
according to the mental rules theory, which of the following is easier: modus ponens or modus tollens?
- modus ponens: easy inference pattern because it requires on one rule
- modus tollens: a somewhat harder inference pattern because it requires using more rules
56
give an example for modus ponens in mental rules theory
- given premise: if mozzi wakes up, then she moves about. mozzi wakes up
- what follows? --> she moves about
57
give an example for modus tollens in mental rules theory
- given premise: if mozzi wakes up, then she moves about. mozzi does not move about
- what follows? --> mozzi did not wake up
58
what is the mental model theory?
a presentation of a possible state of the world. people solve deduction tasks by:
1. creating a set of mental models of the premises
2. taking the set of mental models for each premise and attempting to combine them into a single set
3. checking whether that set of models is consistent with the conclusion
## Footnote
a set of mental models is *like* a truth table, but differs in including only some of the rows (this is why people make mistakes)
59
how does a mental model differ from a truth table?
- "if mozzi wakes up, then she moves about"
- truth table: presents all possible states of the world
- mental model: represents only states of the world where the premises are true
- assumes that peoplease have limited working memory capacity, so they represent as much implicity as possible: full model (w m, -w m, -w -m) vs abbreviated model ([w] m ... where [ ] means this represents all instances of this possibility and ... means there are other implicit models that haven't been fleshed out)
60
give an example for modus ponens in mental models theory
- given premise: if mozzi wakes up, she moves about. mozzi wakes up
- we would have two models:
- premise 1: [w] m ...
- premise 2: w
- when we combine the models, we arrive at only one model that's consistent with the models in both premises: w m
- in this model 'm' is true, so the reasoner concludes that 'mozzi moves about'
- easy because the number of models required is small
61
give an example for modus tollens in mental models theory
- given premise: if mozzi wakes up, she moves about. mozzi does not move about
- we again have two models:
- premise 1: [w] m ...
- premise 2: -m
- these two models can't be combined, so the reasoner might conclude that nothing follows, but with effort, a reasoner could flesh out the first model
- premise 1: w m, -w m, -w -m
- premise 2: -m
- combine: -w -m
- reasoner could conclude 'mozzi did not wake up'
- harder because it requires fleshing out the complete set of models
62
how is the mental model boundedly rational?
| 2 ways
- because working memory limits compromise the quality of mental models
- because we satisfice on effort
63
what is induction?
generalizing from one proposition to another. e.g., X has property P, so Y probably has property P
## Footnote
pretty clear that we are comfortable drawing inferences even when we can't be certain the conclusion if true
64
what is analogy?
drawing connections from one domain to another. e.g., X and P are related in one way, so Y and Q should be related in the same way
## Footnote
sometimes we draw much more indirect inferences
65
what is anduction?
using one proposition to make sense of another. e.g., X is the best explanation of Y, so probably X is true
66
what model is used by induction?
category coverage model: people solve category-based induction tasks by inducing a property from the premise to the conclusions when:
1. the premise and conclusion categories are more similar
2. the premise categories more fully "cover" the smallest category they share with the conclusion category
67
what is the premise monotonicity effect?
adding more categories will always increase the coverage and (total) similarity of the premises to the conclusion
68
what is the premise typicality effect?
robins (typical bird) are more similar to other birds compared to penguins (atypical bird)
69
what is the premise diversity effect?
the smallest category including th epremises and conclusions is mammal. hippos and rhinos only cover a small part of that category, whereas hippos and bats cover a large part of the category
70
what is inclusion fallacy?
robins are more similar to a prototypical bird than to an ostrich. arguably, this is a reasoning error because ostriches are a subset of birds (so if the first argument is correct, the second one should also be correct)
71
what theory is used in analogy?
people use analogies by:
1. mapping a familiar base domain to a less familiar target domain
2. entertaining candidate inferences about target from the base
72
according to the structure mapping theory used in analaogy, what do mappings need to be?
they must be structurally consistent:
- one-to-one correspondence: each elemtn in the base should be mapped to one element in the target
- systematicity: the elements should share matching relations
- candidate inferenes are based on relations NOT attributes
73
our mind contains how many inference systems?
- deduction for drawing valid conclusions
- induction for generalizing from one category to another
- analogy for connecting one domain to another
## Footnote
none of these systems are perfect: deductive inferences limited by working memory and inductive and analogical reasoning both imperfect by nature
74
define probability
P(X) = the probability of X
X is an event or proposition
intuitively, 'P(X) = .5' means: frequentist: "if we drew 100 samples, X would happen 50 out of those times" ... subjectivist: "i am 50% confidfent that X is true"
75
for probability, what are the mathematic propoerties?
- P(X) is always between 0 and 1
- if X and Y are mutually exclusive, P(X or Y) = P(X) + P(Y)
- if X and Y are mutually exclusive and exhaustive, P(X or Y) = P(X) + P(Y) = 1
76
what is bayes' rule?
a rule for updating probabilities based on new information:
- we start with some prior beliefs about hypothesis X: P(X)
- then we see some evidence Y (likelihood: how likely would Y be if X were true?: P(Y|X))
- finally we update our beliefs about X in light of Y using bayes' rule
77
in bayes' rule, what are our new beliefs called?
posteriors
78
in what way are the cognitive tasks inference problems?
- prediction: we have past data and we want future data
- causal learning: we observed data and we want causal relationships
- language: we have stream of sounds and we want meanings
- vision: we have 2d data and we want the 3d world
79
how does the bayesian model of cognition work?
for your inference model, define:
- a hypothesis space: possible answers to the problem
- the evidence: what information bears on the problem
- a likelihood function: to say how well the evidence fits with each hypothesis
- priors: to say how likely each hypothesis was before evidence
## Footnote
then do bayesian inference to obtain the rational posteriors. we use experiments to compare how people solve problems to how an 'ideal' bayesian would solve them
80
how does word learning relate to bayesian process?
- P(Meaning|Evidence) ∝ P (Evidence|Meaning)P(Meaning)
- P(Meaning) captures the bias in our assumptions about word meanings (e.g., whole-objects bias)
- P(Evidence|Meaning) captures how likely the evidence would be under each possible meaning
81
what is coincidence avoidance?
the visual system prefers simpler interpretations of data that require fewer coincidences
82
how do images relate to the baye's rule?
P(Scene|Image) ∝ P(Image|Scene)P(Scene)
83
what are the 3 problems with the bayesian theories?
1. estimating probabilites: how would a bayesian estimate magnitudes?
2. ignoring unlikely events
3. belief updating
84
expand on the problem with estimating probabilities in the bayesian theory
- anchoring bias: when estimating magnitudes, people are influenced by irrelevant numbers
- this is observed in numerous tasks - trivia facts, willingness to pay, imposing monetary penalities, probability judgement
- extensionality principle: propositins that refer to the sme set of events have the same probability, but people assign probabilities to descriptions of events, not events themselves
- subadditivity: X < Y1 + Y2 + Y3
- conjunction fallacy: X > Y
85
expand on the problem with ignoring unlikely events in the bayesian theory
- degrees of belief:
- assigning probabilites to possible worlds
- simulating outcome in each possible world
- integrating info from possible worlds
- digital beliefs:
- assigning probabilities to possible worlds
- simulate just one possible world
86
expand on the problem with belief updating in the bayesian theory
causal imprinting: peole have difficulty "unlearning" causal structures
87
what is resource rationality?
## Footnote
a kind of bounded rationality
- people don't conduct a fully bayesian analysis, but use algorithms that approximate it
- getting into the ballpark is easy (little computational cost) - getting very close to the exact answer is hard (high computational cost)
- a "resource rational" agen balances computation cost against the cost of error
88
briefly explain utility theory
a formal theory of rational decision-making
- doesn't judge people's prefs - it can be rational too
- rather, utility theory is about the consistency of preferences
89
what are axioms?
assumptions/propositions
90
what are the 4 main axioms of utility theory?
1. commensurability
2. transitivity
3. continuity
4. independence
91
what is commensurability?
- everything can be compared
- if A and B are options, then either A>B, B>A, or A~B
92
what is transitivity?
- order of preferences in consistent
- if A>B and B>C, then A>C
- i prefer an apple to an orange and an orange to a lemon; therefore i prefer an apple to a lemon
| utility theory
93
how does intransitive preferences would turn people into money pumps?
| utility theory
- If A > B, then there’s some amount $X so that A ~ B + $X (we’ll assume $1)
- suppose you prefer Apple > Orange and Orange > Lemon, but Lemon > Apple
- igive you an orange; you give me a lemon and $0.50 (since Orange ~ Lemon + $1)
- igive you an apple; you give me an orange and $0.50 (since Apple ~ Orange + $1)
- i give you a lemon; you give me an apple and $0.50 (since Lemon ~ Apple + $1)
- you start and end with a lemon, but you have $1.50 less than when you startef
94
what is continuity?
| utility theory
we can make trade-offs between probability and utility
if A>B>C then there are som e probability p, we can create "mixtures" of A and C so that pA + (1-p)C > B > (1-p)A+pC
e.g., if Apple > Orange > Lemon for me, then we can create lotteries of apples and lemons (e.g., 70% chance of apple, 30% chance of lemon vs 30% chance of apple, 70% chance of lemon) so that I prefer the first lottery to an orange but the second lottery to an apple
## Footnote
this axiom is saying thatthere aren't infinitely steep jumps in preference between options
95
what is independence?
## Footnote
utility theory
- adding a fixed, irrelevant possibility doesn't change preferences
- if A>B, then pA + (1-p)C > pB + (1-p)C
- e.g., if I prefer apples to oranges, then I also prefer a 50/50 lottery of apple/lemon to a 50/50 lottery of orange/lemon
- suppose in our world, you prefer an apple to an orange
- now imagine you instead flipped a coin (heads = apple/orange, tails = lemon)
- the axiom is saying that if the coin flipped heads, you’d still prefer an apple to an orange compared to the world where we didn’t have to flip a coin – why would flipping a coin make you wish you got an orange instead of an apple?
96
if a decision maker satifies the assumptions (in utility theory), then:
1. it is possible to quantify the 'utility' of each option on a continuous scale
2. the decision-maker will choose the option that maximizes their 'expected utility'
## Footnote
utility: the extent to which an option satisfies preferences
expected utility: the utility of each possible outcome weighted by that outcome's probability
97
is commensurability obvious?
| utility theory, revised
- can we compare anything to anything else
- cindy isn't sure whether she wants a career as a lawyer or a violinist (is she indifferent or are they incommensurable?)
- (lawyer, $150,000) ~ (violinist, $75,000)
- (lawyer, $150,005) > (lawyer, $150,000)
- it follows that (lawyer, $150,005) > (violinist, $75,000)
- it seems hard to believe that the extra $5 is what would push Cindy over the top but if it's not, then Cindy is either violating commensurability or transitivity
98
is transitivity obvious?
| utility theory, revised
new studies show that preferences were mostly transitive
- 1 out of 18 participants had some consistent
- violations of transitivity, always for monetary
- rather than non-monetary gambles
99
what is the allais paradox?
individuals often violate the independence axiom by preferring a smaller chance of a larger prize over a sure thing, while simultaneously choosing a sure thing in a separate but related decision, a contradiction in rational choice theory
100
are preferences revealed or constructed?
- revealed: we are acting on a preference that was latent inside our mind (this is what utility theory assumes and most economists believe)
- constructed: preferences don't exist until the decision-maker faces the actual choice, allowing the context to influence the choice (this is what many behavioural economists and psychologists believe)
101
what is preference reversal?
- unproblematic for utility theory if expected utility is different
- but suggests preference construction if there's no difference in expected utility
- response mode (pricing versus choosing)
- evaluation mode (all options at once or one-at-a-time)
- presence of other options (which are not themselves chosen)
102
what is the compatibility effect?
people weigh an attribute more when that attribute is compatible with the response scale
103
what is the pricing task and the choice task?
- pricing task: Pay more attention to dollar amounts
- people price B > A
- choice task: Pay less attention to dollar amounts (and in this case, more attention to probabilities)
- people choose A > B
104
what is the prominence effect?
people weigh the single most attribute more heavily in qualitative tasks (choosing) rather than quantitative tasks (like pricing, rating, or matching)
## Footnote
matching task: Set X1 so that A ~ B
choice task: Choose B > A even when X2 > X1 (implying A > B)
105
what is the evaluability effect?
people weigh an attribute more heavily when it can be readily evaluated
- some attributes are naturally easy to evaluate (e.g., an experience of sound quality)
- knowledge can make an attribute easier to evaluate (e.g., if you knew the possible range of THD)
- attributes can be made more evaluable by presenting options together (joint evaluation) rather than one at a time (separate evaluation)
## Footnote
separate evaluation: THD is hard to evaluate
* B > A
joint evaluation: Sound quality is important and THD can now be evaluated
* A > B
106
what is the decoy effect?
adding a bad option that is easily compared to another option increases the chocie share for the option that is comparable
## Footnote
X vs Y: X ~ Y
X vs Y vs Z: X > Y
107
what is the compromise effect?
adding a more extreme option increases the hcoice share for the middle option
## Footnote
X vs Y: X > Y
X vs Y vs Z: Y > X
108
what is the multiattribute utility theory?
an extension of utility theory that breaks apart options into different attributes
- assign an importance to each attribute
- rate each option on each attribute
- choose the option with the highest attribute ratings (weighted by importance)
## Footnote
this decision strategy is compensatory - higher values on one attrubute can "compesate" for lower values on another
109
in decision strategies for utility theory, what is averaging?
choosing the option with the highest weighted attribute ratings. this strategy is computationally very demanding
110
in decision strategies for utility theory, what is conjunctive?
set minimu cut-offs for each attribute and only consider option that meet all cut-offs. the elimination stage of this strategy is non-compensatory
111
in decision strategies for utility theory, what is elimination by aspects?
rank attributes and set cut-offs for each. for first attribute, eliminate any options that fail to meet threshold, continuing until one option remains.
112
what are non-compensatory decision strategies like?
- sophisticated version of satisficing (choosing a good enough option)
- for most real-world decisions, a full MAUT analysis would require too much working memory to carry out without external aides (e.g., spreadsheets)
- at best, we can be boundedly rational in complex multi-attribute decisions
- people can flexibly select decision strategies, investing more effort when the decision has higher stakes
- another example of resource rationality - balancing the costs of computation against the costs of error
113
what is diminishing marginal utility?
- each additional unit brings less utility than the previous unit
- you get a lot of satisfaction from your first apple, but by the time you're on your second, you're partly satiated
- with your first $500 you buy all the things you really need - and then the next $500 goes to things that are somewhat less important, and so on
114
for gains/losses, what are people more likely to be?
- risk-averse for gains
- risk-seeking for losses
## Footnote
another example of a preference reversal (based on how the options are framed – gain versus loss)
115
what is the prospect theory?
a theory of decision-making under risk which assumes that:
1. preferences are reference-dependent (i.e., depends on your starting point)
2. rather than diminishing marginal utility for wealth, we experience diminishing sensitivity to both gains and losses
3. people are loss averse: losses loom larger than gains (~2 times larger)
4. people overweight small probabilities and underweight large probabilities
116
what is possibility effect?
paying a premium to make a good event possible or a bad event impossible
117
explain prospect theory in terms of bounded rationality
no current consensus
- we pay more attention to losses because the cost of ignoring them is higher than the cost of ignoring gains
- ”efficient coding” of magnitudes requires reference dependence (like adaptation for light levels by the perceptual system)
- we represent magnitudes in terms of their rank within a distribution, not their absolute level – in the real world, there are lots of small losses and fewer large gains
117
what is the certainty effect?
paying a premium to make a good event certain or avoid a bad event being certain (rather than very likely)
118
what is game theory?
a formal approach to studying human interaction that assumes:
1. people are rational (selfish) utility-maximizers
2. people assume that other people are rational utility maximizers
## Footnote
generalization of expected utility theory to situations with multiple players, each trying to maximize their own utility given what others will do
119
what is normal-form in game theory?
this matrix format for representing a game
120
what is simultaneous game in game theory?
a game in which both players choose the at the same time
121
what is the dominant strategy?
a strategy that has a better payoff no matter what the other player chooses
122
what us the nash equilibrium?
an outcome where no player could do better by unilaterally deviating (i.e., holding other players' choices constant)
123
in game theory, what is extensive-form?
this tree format for representing a game
124
in game theory, what is a sequential game?
a game in which players choose one at a time
125
what is backward induction?
start at the last branch of the tree and work backwards
126
what is the cooperative dilemma?
players cna maximize total payoff by cooperating, but each player has an incentive to defect
## Footnote
classic example is the prisoner's dilemma
127
what is the coodination dilemma?
players have aligned interests but are uncertain what the other player will do
## Footnote
classic example of the stag game
128
what is the bargaining dilemma?
players are dividing a resource but are conflicted over how to distribute it
## Footnote
classic example is the ultimatum game
129
but sometimes we do cooperate - modern world runs on widespread cooperation. why?
## Footnote
three ways of changing the payoff structure
- favorable strategic environments --> Repetition, Punishment
- cognitive adaptations (biological evolution) --> Emotion
- social adaptations (cultural evolution) --> Institutions
130
what is agent based models?
simulations that examine how collecitve behaviour is affected by individual decisions
- create a set of agents
- specify how individual agents behave
- let the agents interact with each other according to the set of rules
- measure emergent properties of the simulated group
131
what are the examples for the rules for repetition (cooperation)?
- tit-for-tat: cooperate on the first game, and copy the other player's move from the previous game (conditionally cooperate)
- tit-for-tat + cheating (JOSS): same as tit-for-tat but defects 10% of the time after the other player cooperates
- tit-for-tat + escalation (SHUBIK): smae as tit-for-tst but punished the second defection with 2 defections and the third with 3 defections, etc.
| best strat is tit-for-tat
132
what are the properties of best-performing strategies in repetition?
- niceness: not the first to defect (all the top strategies werenice, all the bottom were not)
- forgiveness: eventually cooperating again after a defection
- the best strat bepends on what other strats are in the tournament
## Footnote
some poor-performing strats were "kingmakers" - their compatibility with different nice strats helped determine a winner
133
what is nash equilibria?
a stable state where each player in a game chooses the strategy that is their best response to the other players' strategies
134
give some examples of real-world situations that are coordination dilemmas
- technology standards: etter if we can all agree on one shared standard (e.g., USB-C)
- language conventions: language is most useful if we can all agree on shared vocabulary
- political activism: protesting a dictatorial regime benefits everyone, but only if enough people show up
135
how do we coordinate?
facilitated by common knowledge - just private knowledge isn't enough
136
what mechanisms generate common knowledge?
- communication: we use language to announce our intentions in the presence of another
- conventions: everyon drives on the right side of the road
- focal points: one outcome is more salient or obvious
137
give real-world situations that are bargaining dilemmas
- economic transactions: bargaining over wages or haggling over prices
- legal negotiations: divorse settlements, dividing an inheritance
- international relations: everyone benefits from free trade agreements but each party wants specific provisions
- spousal relations: allocating household chores
138
what are commitment devices?
mechanisms voluntarily adopted now that remove future options
139
why do we have commitment devices?
- because we think our prefs might change (e.g., alarm clocks or pensions)
- because it changes others' behaviour (e.g., encourages long-term coordination --> binding contracts or wedding rings)
140
what do commitment devices require?
formal institutions or innovations like governments, courts, tech, etc.
141
in what way are emotions commitment devices?
- they commit us to curses of action that are outside our short-term interest, but by doing so precommit us to courses of action
- for this to hold: emotions must be involuntary, must be knowable by the partner
- arguably this is why facial expressions evolved
142
where did collective behaviour emerge?
in many domains, group properties are not merely the sum of their parts (are water molecules wet? what colour are gold molecules? what does an individual neuron in your amygdala think about the election?)
## Footnote
in human societies too, many important properties emerge at the level of collectives
143
what does the Schelling model tell us about collective behaviour?
- shows there is top-down discrimination or individual prejudice
- high degree of segregation can occur even when individuals have even a pretty small preference to be around similar people (e.g., they want 25% of their neighbours to be in the same group)
- key point: surprising collective behaviours can emerge even when no individual intends them
144
how is it that we can do more in groups than we can individually?
- larger groups --> greater division of labour and more ideas
- division of labour: dividing up a task into smaller pieces allows workers to be more productive
- workers specialize doing what they are naturally or trained to be good at and trade with other specialists
- ideas: innovation increases the productivity of each individual, larger groups are able to get better faster
145
what other mechanisms have evolved for cooperation?
- emotions like guilt and shame
- shared genes --> kin selection (family)
- repeated interactions --> reciprocal altruism (friends)
- common knowledge --> reputation (small groups)
## Footnote
these mechs don't work in large groups = interactions are often one-shot and anonymous
146
in collective behaviour, what are institutions?
the “rules of the game” that structure human interaction within societies:
- they alter incentives and provide predictability
- formal institutions are cultural products. they arise through cultural evolution and are maintained, altered, or eroded across generations
## Footnote
e.g., economic, legal, political, epistemic
147
what is the tragedy of the commons?
individually rational behaviour leads to overuse and eventual depletion of a shared resource
148
how do institutions solve the tragedy of the commons?
- property rights
- the problem is that the resource is collective – we each have an incentive to exploit it as much as possible
- property rights create incentives for stewardship
- people have to come up with ways to use property rights to solve common dilemmas: fishing rights, patents and copyrights, radio spectrum auction
149
places with ____ property rights are ____
places with stronger property rights protections are (much) richer
150
what is the expected utility calculation?
EU = U(A)xP(A) + U(B)xP(B) + U(C)xP(C) + ...
PSYCH 256 - Cognitive Science
flashcards
Decks in class (4)
# Cards
