CP

• Making an assumption based on observation

• Assumption: choices reveal latent preferences

  • If Anna chooses an apple over an orange, Anna prefers the apple

• p = Player 1 chooses dodging (probability)

• 1 - p = probability of player 1 not choosing dodging (straight)

• q = probability of Player 2 choosing dodging

• 1 - q = probability of player 2 not choosing dodging (straight)

  -> depending on the probability of behaviours of other player, another action has a higher expected value

  -> mixed nash equilibria try to find probability point where EV is the same

• EV(p) -> Expecte value of Player 1 dodging, dependent on probability of behaviour of Player 2

  • EV(p) = 0 x q (-> points action will get x probability of other player showing matching behaviour with points) + 1-q x -2 = 2q - 2

  • EV(1 - p) = q x 2 + -5(1-q) = 7q - 5

• Set them next to eachother to find point where q leads to same value:

  • 2q - 2 = 7q -5

  • 5q = 3 -> q = 3/5

• If theres a possibility of P2 dodging of 60%, player 1 is indifferent between dodging and straight (same EV)

• If the probability of player 2 dodging is aboive 60%, EV of Player 1 to go straight is bigger

• For player 2 -> same

  -> draw lines

• Questions asked:

  • How to spend money

  • How to allocate effort

  • How to plan for the future

• Answered by decision theory, which is used by

  • Economics: How to make “optimal” choice

  • Psychology: How people actually make choices

Which leads to consequences for social policies

• Theory on how people SHOULD make decisions (rationally)

• WARP: Weak axiom of revealed preferece/Completeness assumption

  • When given the choice between A and B, we assume that the consumer prefers

    • A

    • B

    • Neiter/No preference

  • Violation of WARP: Consumer prefers both A and B

• Independence of irrelevant alternatives assumption (IIA):

  • Assumption: If someone prefers A over B, this preference will hold up consistently in every circumstance

  • Violation: If someone prefers A over B, but will choose B if they have to choose between A and a grape or B and a grape

• Transitivity Assumption

  • Assumption: Concluding that if A > B, and B > C, then A > C

  • Violation: the money pump thought experiment

    • If object ranking is not rational, a person would want to trade B for A, A for C and C for B -> this is how people can be exploited

      • Being rational is in ones own best interest, since being irrational opens the gates for being exploited

Choice patterns which satisfy simple, logical assumptions of choice

• Decisions reveal preferences

• Preferences can be ranked/ordered from low to high

• Decisions can be characterized as if the decision maker would maximize their “utility”

• Property in any object that

  • produces benefit, good, wellbeing, happiness, advatage

    • and/or

  • Prevents the happening of mischief, pain, evil, unhappiness

• to the party whose interest is considered

• If we can quantify the pain and pleasuer of decisions, we should choose those decisions which maximize the total expected utility of everyone if affects (maximize pleasure, minimize pain)

• The trolley problem

• According to utilitarianism, we should always choose the option which saves most lives (even by sacrificing one)

• People struggle to act accordingly when they have to actively push someone to their death

• Trolley-dilemma is being used for self-driving cars to decide who should die in case of an accident

• Problems:

  • There are preferences based on age, occupation, animal vs. humans etc.

  • Those preferences are further influenced by culture, i.e. they differ between west/south/east

    -> should cars decide depending on the location where they drive?

• Calculated by multiplying each possible outcome by its likelihood and then summing it together to make a rational decision under risk

• Example:

  • Option A: 50% chance of winning 100.-, 50% chance of winning nothing

  • Option B: Getting 40.- for sure

  • 0.5 x 100 + 0.5 x 0 = 50.- -> 50.- > 40.- -> Rational choice would be choosing option A (expected value optimiser)

• Paradox in relation to expected value

• Choice: a fair coin is tossed

  • If head appears, you win 2 Euros and the game goes on

  • If head appears, you win 4 Euros and the game goes on

  • If head appears, you win 8 Euros and the game goes on

  • etc.

  • The game ends when tails appears

  • How much are you willing to pay?

• Calculating with expected value, it would look like this:

  • .05 x 2 + 0.25 x 4 + 0.125 x 8 etc. -> infinite

  • Accordingly you should be willing to pay an infinite amount of money to play (according to expected value maximizer)

  • Yet, almost no human being is willing to do that

• Possible explanation

  • The value of an item is not based on the price, but rather the utility it yields (i.e. winning 1000.- is more relevant to a poor person than to a rich person, even though it is the same amount of money objectively)

    -> Money chages its value depending on who has it

• Utility: depends on psychology/perception (contrary to value)

• Relationship between utility and value is NOT (necessarily) linear

  -> Chages things around decision making under risk

• Alterantive to the expected value theory/calculation

• Instead of including the actual value, the individual utility is inserted

• I.e. if the utility of earning 40.- is 30, we can then calculate the utility of winning 100 and 0 CHF -> can change the result of the best “rational” decision

  • The difference in utility between 40.- and 100.- might be smaller than the actual difference in value

-> Maximizing utility (rather than value) is also rational, since rationality = sticking to your preferences, being consistent

• Risk aversion: Concave utility function (value > utility)

• Risk neutrality: Linear utility function (value = utility)

• Risk-seeking: Convex utility function (value < utility)

• Risk preferences are not considered irrational (as long as chooser consistently maximises utility)

• There is a context depencence to risk preferences (otherwise risk adverse people would never go to a casino)

• Individual risk preferences can be assessed by presenting different decision problems and always adapting option A (the risky decision) -> depending on the decision pattern you can tell the risk preference of that person

• On average across groups people are rather risk-averse (concave curve)

  • however, not to an extreme, they are around 0.5 risk averse

• People discount value over time by a certain factor

• Can be calculated with the following equation:

  • U(X) = X x e^-kD

    • U(X) = utility of X

    • X = actual value of X (i.e. 20.-)

    • k = discount factor

    • D = delay (i.e 5 days)

• Value of X changes/diminished depending on the Delay and the k factor

• Time preferences are also rational, as long as they are dynamically consistent

• Through choice patterns, the latent time variable k can be made visible

• When participants are asked whether they would prefer

  • 3400 Dollars now

  • 3800 Dollars in a month

• People in richer countries are far more likely to choose the 3800 Dollar option than people from poorer countries

• Inflation (money might lose it’s worth)

• Interest rate over money

• Uncertainity about future wellbeing

• Policies should be aimed at making decision consequences transparent (perfect information)

• No need to help decision makers make decisions

• People would know best what they want

• Libertarian principle of governance

• decision makers have perfect memory

• Decision makers percieve and integrate all relevant information

  • therefore if a person goes grocery shopping, they would be aware of all products in the shop (including the relevant information) and make the most rational decision based on that knowledge

• Also “the certainity effect”

• Operationalisation

  • Two Rounds in which people get to choose between

    • Round 1:

      • Option A1: 100% chance of winning 1 Million

      • Option B1: 89% Chance of winning 1 Million, 10% chance of winning 5 Million, 1% chance of winning Notion

    • Round 2:

      • Option A2: 11% Chance of winning 1 Million, 89% chance of winning nothing

      • Option B2: 10% chance of winning 5 Million, 90% chance of winning Nothing

• Rational choice: choosing A1 and A2 or B1 and B2

  • In both Rounds, Option B has 1% higher chance of winning nothing

  • In both rounds, Option B has the better expected value (but in Round 1 it is more risky)

  • Reality: Most people chooose 1A and 2B

  • This does not correspond to the indepencence assumption!

• Possible explanation:

  • 2A and 2B are both unlikely, but 2B has higher EV

  • in Round 1, people might prefer the certainity of option 1A

• Operationalisation:

  • Setting: there are two urns

    • Urn A has 50% blue balls, 50% red balls

    • Urn B also has exclusively red and blue balls, but nothing is known about the distribution

  • There are four bets people have to choose between:

    • Bet 1A: win 1 CHF if a red ball is drawn from Urn A, 0 CHF otherwise

    • Bet 2A: win 1 CHF if a blue ball is drawn from Urn A, 0 CHF otherwise

    • Bet 1B: win 1 CHF if a red ball is drawn from Urn B, 0 CHF otherwise

    • Bet 2B: win 1 CHF if a blue ball is drawn from Urn B, 0 CHF otherwise

  • Peoples preferences:

    • 1A = 2A

    • 1B = 2B

    • 1A > 1B

    • 2A > 2B

  • Problem: it doesn’t make sense for people to both prefer 1A to 1B and 2A to 2B, since there can’t be both more blue balls AND more red balls (proportionally) in Urn A than in urn B

    -> People believe that Urn A has both more blue balls than red balls and more red balls than blue balls! -> Contradiction

• Possible psychological explanation:

  • Ambiguity aversion

    • People prefer situation where they can attach (certainly) probabilities to outcomes

• Two kinds of frames:

  • Gain frames: focus on what can be won/saved

  • Loss frames: focus on what can be lost

• People tend to choose certainity when something is presented in a gain frame and risk when it is presented in a loss frame

  • therefore people would choose to get 240 Dollars for sure instead of a chance of 25% at winning 1000 Dollars (and 75% chance of not winning anything) but would also choose to risk losing 1000% with a 75% chance and not losing anything with a 25% chance instead of losing 240 Dollars for sure

• This means that…

  • Peoples risk aversion/risk seeking behaviour changes depending on framing

  • The completeness assumption/weak axiom of revealed preferences (i.e. A > B means Person will prefer A over B in every situation) is violated

  • Same consequences lead to different decisions (depending on framing)

• Theory by Tversky and Kahnemann

• Descriptive theory of decision making

• Alternative theory to Expected Utility theory, attempt at integrating the framing effect into theory

• Creates a curve (with value in the y axe and outcome in the x axe) which is steeper in the losses department than in the gains department

• Criticism: it’s not surprising that adding more parameters makes a prediction more specific

• Expected Utility theory:

  • EU(x) = p x U(x)

    • P = probability of getting something

    • U(x) = utility of x

  • U(x) = x^a

    • Alpha = determines the shape of the utility function (concave/convex) -> higher Alpha = more risk averse/more concave curve

    -> EU(x) = p x x^a

• Prospect theory:

  • same function as prospect theory, BUT:

    • In gain frame/ if x > 0, then v(x) = x^a

    • In loss frame/ if x < 0, then

      • v(x) = -Lambda(-x^a)

        • -x^a = mirror of gain frame, so concave function becomes more convex

• Value is not value in absolute terms, but rather distinction between losses and gains

  • Degree of riskaversion/riskseeking governed by Alpha (0 < Alpha < 1)

  • Losses loom larger than gains governed by Lambda (1 =/< Lambda)

• Point at which peoples perception as a gain changes into perception of something as a loss (and vice versa)

• This point is (to some degree) subjective and also depends on framing

• Influences, which equation of prospect theory is used

• Manipulation fo the reference point

• A seller sees more value in the object he/she sells than the buyer sees in the object they may aquire

  • When people are randomly put in the seller/buyer role and given a random object, the sellers ask for more money for it than the buyers are willing to give (even though its the exact same object)

• Different psychological explanations:

  • Loss aversion: Buyer might be averse to parting with something they own

  • Framing effect: Buyer sees selling thing as a loss, buyer sees it as a gain -> losses are valued higher than gains

  • Mere ownership effect: if a person owns something, they value it as better solely because of the fact that they own it

  • Status quo bias: preference for the current status of affairs

• Part of prospect theory

• People participate in a game weherein…

  • In stage 1, they have a 75% chance of ending the game without winning anything and a 25% chance to move into the second stage

  • In the second stage, you can choose between

    • Option A: 30 Dollars with p = 1 (chosen by 74%)

    • 45 Dollars with p =0.8 (chosen by 26%)

• People treat option A like a cerain option, but they don’t factor in the initial 25% chance of even getting there -> pseudo-certainity effect

• People focus on the “certain” effect in later stages of a decision while ignoring the fact that reaching that stage is conditional -> leads people to act as if outcome is more likely than it is

  -> option b has higher expected value

• Expected utility theory (EUT) draws a straight, diagonal line between subjectives and objective probability

• According to prospect theory

  • Small probabilities are overestimated

  • High probabilites are underestimated

• Phenomenon wherein a new, inferior option shifts preferences toward one of the original options by making it seem comparatively better

• Specifically, if you have to choose between

  • 85 Dollars for online subscription

  • 125 Dollars for print & online subscription

• and most people were going to choose the 85 Dollar option, their choice might shift if a third, worse option is added close to the print option

  • 125 Dollars for only print -> leads more people to choose the 125 print & online version, because it is comparatively significantly better!

• If an added optionC is strictly dominated by an already present option A, it increases the likelyhood of the dominating option being chosen (over a completely different option)

• Explanation: A is easier to compare to C than A to B -> A is clearly better than C -> A as the best option (even between A and B)

• Violation of independece of irrelevant alternatives (IIA)

• Phenomena wherein the percieved value of an object/many objects is adjusted based on the relative values of all other available options

  1. Options “compete” with each other

  2. When another option is introduced, all options are reevaluated and “normalized”

• In the lab, if people had to choose between two snacks, one with subjective utility of 9 and another with a su of 10, their choices were more likely to be “wrong” (i.e. choosing the 9 over the 10) if a kind of decoy snack was introduced whose value was close to them

  • So a decoy of 6 was less likely to lead people to choose the 9 over the 10 than a decoy of 8

• Neurocognitive explanation:

  • Neurons are limited

  • if distractor is introduced, some firing has to be used for that -> we split up the total amount of spikes, thus making it harder to distinguish

• Normally, if a person prefers to buy A instead of B, it means that

  • value(A) - Price(A) > Value(B) - Price (B)

• If we were to detract the same amount of price on both sides, the preference should remain the same

• However, this changes if the detraction of the same amount of price leads to one of the two prices being zero

  • i.e. if Originally a was 10.- and B was 2.- (and we preferred A), detracting 2.- from both should lead to A still being preferred

  • Instead, because the prize for B is now zero, people tend to prefer B

    -> zero pricing of a good is not only seen as a decrease in cost, but as an added value (because otherwise people would still prefer A)

• In general, people act as if the value of things decreases with time (time discounting)

• However, there seems to be an added effect for the present day

• If people are to choose between 50.- in three months or 100.- in six months, they will choose the 100.- option

• However, if people have to choose between 50.- now or 100.- in three months, they will choose the 50.- now, even though it is the same time distance

  -> this would lead the person in the first choice to regret their choice three months later -> inconsistent & exploitable

• exponential discounting does not account for that (since the value should decrease consistently at the same rate for a three months difference) -> hyperbolic discounting

• Gains are discounted more than losses -> People prefer to get gains now and losses later

• Dynamically inconsistent choices

• Present day bias

• Impulsivity/self-control issues

• Regret (you will regret the choices you made in the past because the value changes as you inch closer to now)

• Definition

  • Simple judgement or decision rule

  • Is fast but can lead to

    • Suboptimal decisions

    • Wrong judgements

    = non-optimal approximation

• Used when…

  • Information is incomplete/uncertain

  • Information is too much

  • Information is not perfectly relevant

    -> the alternative would be increasing the information quality (i.e. gather more information) but that is costly (takes time and requires energy)

• A way of manipulating judgements

• When people are given or asked to produce a random anchor (for instance a random number) and then asked to estimate something (for instance a date), they are going to estimate a data close to the randomly produced number (although they are aware that the number is random and not related to the date)

• People tend to adjust the number/shift it away from the anchor because they are aware the anchor is not a good reference point, but they usually don’t adjust it sufficiently

• Anchoring influence is hard to override, even if clearly non-informative

• judgement based on available heuristics

• We assume things that we can think of more easily are more frequent

• Pathway

  • Uncertainity -> retrieving samples from memory -> more words/memories of a certain group available -> judgement based on availability

• the conjunction fallacy

  • When we think that something is very representative of a certain category (i.e. an uptight person as representative of being an accountant), we assume that it is more likely that this person belongs to that specific category AND also has a characteristic which doesn’t fit said category (i.e. plays jazz) that that this person has exclusively another characteristic which does not fullfill the expected category

  • This does, however, not make sense, because it is more likely that someone/something belongs exclusively to A or exclusively to B rather than that the person is at the crossover between A and B (because in total this is simply a smaller amount)

    • Also, someone cannot be both A and B and not belong to A or B, but someone might belong to A or B but not to both A and B

• judging randomness

  • If for a coin the number side is A and the face side is B and the coin is tossed 6 times, people assume that

    • A B B A B A is more likely than

    • A A A B B B

  • because the first sequence is more representative of randomness

  • But actually, they both have the same probability (1.6%)

• Betting on the future

  • If coin has been tossed four times and it has shown head all those times, people assume that it is more likely for the next toss to be number

  • Indeed, throwing head five times has a small probability

    • P(HHHHH) = 1/2 x 1/2 x +1/2 x 1/2 x 1/2 = 0.03125

    • However, the overall chance of throwing head is 0.5, and the coin does not remember what has been tossed before -> P(H I HHHHH)

    • The probability of tossing 4 times head and then number is also 0.03125 -> this is about the probability of predicting a specific order (and includes the aforementioned tosses, i.e. it is so low because of the many different variations of order there could be with five tosses and two sides (i.e. HHHHH, HHHHN, HNHHN etc.)

    • Because those probabilities are mistaken, people would be more willing to bet more on number than head for the next toss, even though they both have the same probability!

      -> this also happens when people have to guess the sex of the next child

• Leaning bias of uncontrollable events

  • When people get close to winning something, for instance with a slot, they assume that winning is now more likely, even though the spinning of the slot is completely random

  • Also smiliar to the betting on the future thing

• Operationalisation

  • People have to operate a slot where they win something when they get two of the same symbols next to each other

  • In some trials the people can only press the start, in other they can alos press when the slot has to stop (but it is so fast it is kind of uncontrollable)

  • Two kind of trials:

    • Full miss trials: symbols are not the same and the corresponding same symbol is far apart

    • Near miss trial: symbols are not the same, but they are close to each other (like one position apart)

  • When people are then asked if they want to continue the game, they…

    • Are more likely to want to continue when it was a near rather than a full miss trial, but only when they are the ones who can stop the slot (likelyhood of full miss trial pretty much the same, even less likely to want to continue in a near miss trial when controlled by a computer)

      -> Problem: the game is completely casual and does not remember what happened before, so a near miss trial does not make the next trial any more likely to be a winning trial than a full miss trial!

• Why are people more likely to want to continue in a near miss trial?

  • People treat uncontrollable events as if they were controllable

  • Thus, they see a near miss as being close to winning/learning how it works -> kind of like when we learn to bike and we manage to drive a few meters and then fall, as this would signal to us that we’re close to mastering it

  • This is also supported neurologically:

    • more activity in ventral striatum in near miss trials (subcortical structure assoiciated with reward-based learning)

      • In a learning task it would make sense for the brain to reward us for near misses, as this motivates us to keep going and fully mastering it

    • The brain thus treatsit like a learning task (and thus controllable), although it is uncontrollable

• I.e. Misjudging conditional probabilities

  • If there is an illness, and incidence rate (amount of people sick right now) is 1%

  • The probability of getting a positive result from a test if someone is sick is 100%

    • p(P I s) = 1.00

  • The probability of getting a positive result from a test if someone is healthy is 5%

    • p(P I h) = 0.05

  • When people are asked to estimate how likely they are to be sick when they get a positive test result, they think it’s about 95%, thus neglecting the baserate

    • If there are more healthy people getting a tets than sick people (which is likely the case with an incidence of 1% sick people), the toatl amount (and thus also the relational amount) of people who are healthy and get a positive result is bigger than the amount of people who are sick and get a positive result -> false positive more likely than positive positive

      -> people neglect the incidence rate when estimating probabilities

    
    

• P(P) = p(P I s) x p(s) + p(P I h) x p(h)

  • P(P) = Probability of positive test result

  • p(P I s) = probability of getting a positive test result when sick

  • p(s) = probability of being sick in general (incidence rate)

  • p(P I h) = probability of getting a positive test result when healthy

  • p(h) = probabiliy of being healthy in general (incidence rate)

• P(s I P) = (p(P I s) x p(s))/p(P)

  -> Probability of being sick when getting a positive test result = probability of getting a positive result when sick x probabilty of being sick in general (incidence rate) divided (together) by the total probability of getting a positive test result

  • p(P) = p(P I s) x p(s) + p(P I h) x p(h)

    -> if the incidence changes then the division chnages -> probability of being truly sick when getting a positive test result changes

• When faced with multiattribute decisions, people are overwhelmed

  • i.e. When having to choose a new laptop, you have to pay attention to different brands and websites and features and prices and…

• Problem: we want to make the best choice (maximizing) but we also don’t want to use too many resources like time, money etc. (satisficing)

  • Assumption: information search and processing is

    • sequential

    • costly

• Satisficing heuristics: searching through alternatives until an accepatble option is found

• Satisficing can be seen as an ecologically rational strategy because it gives the optimal tradeoff given the constraints

  • Ideal: search cost, time cost should be minimized, satisfaction with chosen option should be maximized

  • However, satisficing can lead to non-optimal choices (i.e. neglecting an option which would be better)

• Observation:

  • When people are offered too many choices, they are less likely to make a decision/choose/buy one than if there are fewer options to choose from

  • However, people do not tend to “try”/seek information about more options in the more option category than in the less options category

• Expectation:

  • This is not what would be expected, since if there are more choices, there’s a higher likelihood of there being an option which fits ones desires

• Possible explanation:

  • Too many choices are overwhelming (thus information for everything is not gathered)

  • This increases regret aversion -> Person might not have tried the best option, thus making a decision now could lead to wrong decision

• NOTE: recent studies have found little evidence for the choice overload effect

• People tend to value (i.e. be willing to pay more for) same thing more if it is…

  • Pictured rather than written (text condition)

  • Physically present rather than pictured

  • In front of someone without a plexiglas vs. behind a plexiglas

  • Behind plexiglas instead of on a picture

• Systematic judgement errors and violations of rational choice principles make decision makers systematically exploitable, which is an issue in terms of…

  • Marketing/advertising: Consumption behaviour can be manipulated

  • Paternalism: Restricting liberty or autonomy to help decision makers to avoid making mistakes (i.e. we can change the decision architecture such as gain/loss frame to help people make the right decisions)

  • Nudging: manipulating consumer behaviour, health behaviour, financial decision making etc.

• It also leads us to rethink certain policy implications, i.e. setting incentives/punishing people for crimes is based on the assumption of a rational consumer who calculate the costs and benefits correctly and behave based on those decisions (mechanism design), but this is not the case

• Popcorn

  • Small: 6$

  • Medium: 10.50$

  • Large: 12$

    -> Decoy effect

• 100 Nights/days free trial! -> endowment effect

• Experience stores such as the app store

  -> endowment effect, reachability effect

• Writing in a parking lot that if a car door opens you’ll have an accident with your bike and you’ll be happy to have insurance

  -> availability heuristic

• “Buy now, pay later!”

  -> hyberbolic discounting

• Exact percentages of how many wrinkles a cream reduces etc.

  -> avoiding ambiguity aversion

• Asking if you’re sure you want to cancel and if you don’t want to check out certain shows before leaving when cancelling a Netflix subscription

  -> triggering loss aversion

• Free delivery/buy one, get one free!

  -> zero price effect

• Showing the new price of 20$ and the crossed out price of 80$

  -> anchoring

• Recognition heuristic: if one of two things is recognized, decision makers judge the recognized object to have higher value

  • Hence why logos are often very simple

• = restricting choices and actions based on what is “the best” for people

  • Compulsory seat-belts

  • No smoking places

  • Corona pandemic

• Typical evolution

  1. Restriction is introduced

  2. Psychological reactance: people resist it, are unhappy about being forced to do something

  3. Status quo changes

  4. Increased acceptance over time as changed norm becomes new norm

• = influencing choice without restricting it

• i.e. Taking the knowledge about choicepatterns we have from psychology and introducing it into framing so that people are more likely to make the choices that are best (for them or whomever introduced the nudging)

  • People are pushed to make better choices

• Ways of nudging:

  • Changing the status quo

    • Status quo bias: people are more likely to leave things as they are than to change them

    • If avoiding a desirable behaviour requires actively changing something (instead of the other way around), people are more likely to engage in/accept the desirable behaviour

    • Changing the status quo does not restrict choice set, but can significantly impact choice (default effect)

    • Examples

      • Opt-in vs. opt-out for organ donation: having to actively sign up to become an organ donor vs. having to actively ask to be dismissed as an organ donor (the second leads to way more people being subscribed to donating their organs)

      • Introducing a renewable default option for energy (even though it costs more) where people would have to actively unsubscribe to get rid of it -> more people consume renewable energy

  • Nudging though loss frame

  • Nudging through information:

    • making information easier to process increases the probability that people follow the recommendation (i.e. nutriscore)

    • Avoiding base-rate neglect by presenting frequencies

      • instead of presenting probabilities (%) which are hard to understand, one could present frequencies (50 out of 1000 people instead of 5%)

  • Commitment devices to avoid hyperbolic discounting

    • i.e committing to go to the gym starting january 1st -> if person does not follow through, they’ll be forced to donate to charity they don’t like

      • commit to future costs/punishment if plan fails

• Preferne for the current state

  • related to

    • endowment effect

    • loss aversion

    • Reference-dependent choice (prospect theory)

      
      

• When you want people to sign up for an event (i.e. science conference) early, you can use a gain or a loss frame:

  • Gain frame: Sign up early to get tickets at the discounted price of 120$ instead of 150$!”

  • Loss frame: “If you sign up too late you have to pay an additional fee of 30$ next to your 120$”

    -> same difference in terms of money, but people are more likely to sign up when it is presented in the loss frame

• Pro

  • Easier to process information like that, so people might pay more attention to it

• Contra

  • Misunderstanding: green label might be seen as very healthy in general

  • Psychological licensing: consuming green products allows consumers to consume red production

• Opponents

  • Nudging manipulates people

    • contrary to our idea of self-determination and dignity

  • People need to be educated rather than nudged or societies should democratically decide which behaviour to restrict

  • Who decides what are good nudges?

    • Risk of abuse of power/undemocratic

  • Nudging can reduce the trust in governments

    • feeling of being manipulated

• Proponents

  • Nudging can safe lives

  • Choice architecture always have defaults (never “neutral), nudging just changes the defaults for the better

  • Some choice anomalies are like visual illusions -> they cannot be overriden or unlearned

  • Nudges do not restrict choices

• Humans seen as bounded rational

• Rational choice models need to be revised if the aim is to derive descriptively more valid decision models that are better to predict choice (e.g. prospect theory)

• Problem: too many free parameters

  • arbitrary models

  • arbitrary number of heuristics

• Our brain is constrained

  • number of neurons

  • Energy consumption

• Our brain tries to process information as efficiently as possible given those constrains

  -> Optimization necessarily leads to shortcomings

• Heuristics are adaptive and ecologically rational

  • Those heuristics exist for real life problems we often encounter(ed) and make sense there

  • Doesn’t make sense to see humans as irrational based on this

• Decision making should be understood as a collection of adaptive strategies rather than as a unviersal strategy that are judged against formal logic or probability theory

• 1. Objective/observable variables:

  • risk (at beginning of model)

  • Outcome (at beginning of model)

1. Based on the risk and outcome, a cognitive evaluation is made (subjective/latent/psychological)

   1. This is influenced by anticipated emotions

2. Based on the cognitive evaluation, the subjective risk and value are estimated (subjective/latent)

   1. This is influenced by the emotions/background mood

3. This leads to an observable, objective choice and outcome

   1. The outcome leads to experiencing certain emotions

• Affect heuristics: reliance on feelings to guide decision

  • i.e. “I have a gut feeling this might be a bad idea”

• Made up of…

  • Affect-as-information theory

    • Emotions are evaluative (good/bad)

    • Emotions are used as information, i.e. if I feel scared, theres something bad about this choice

    • Emotions are treated as information even when they’re not completely clear

  • Risk-as-feeling hypothesis

    • Emotions bias cognitive evaluations

      • Risk assessment is not really rational but rather based on feelings

  -> this leads to the premise that emotions influence decisions in predictable directions

• Part of affect heuristics

• When strong emotions are involved, people tend to focus way more on the badness of the outcome than on the probability of something happening -> hence why people are more scared of terrorism than car accidents, although terrorism is way less likely than car accidents

• Avoiding information which could lead to negative affect

• i.e not wanting to know whether one has the predisposition for an illness, because….

  • If the person does NOT have that predisposition, the resulting relief is small (if they never thought about whether they have that illness before)

  • If the person DOES have that predisposition, the resultig affect will be very negative

    -> better to avoid information alltogether

• U(not knowing) > U(knowing I negative) + U(knowing I positive)

• Also seen in games where you can win/lose something:

  • The higher the probability of winning, the more people want to know the result beforehand (even though they are going to get the reward anyways)

  • The higher the probability of losing, the more people do NOT want to know the result beforehand (even though they are going to lose anyways)

    
    

• People might avoid important information (i.e. to stop smoking for health reasons) to avoid negative affect

• Possible solution: combine with humor (positive affect)

• Avoid negative emotions

  • i.e. Avoid knowing whether one has predisposition for illness

    -> regret aversion

• Maintain surprise and suspense

  • i.e. Not knowing the gender of a baby

• Gain strategic advantage

  • i.e. avoid liability and responsibility

• Maintain fairness and impartiality

  • i.e. Not knowing gender of applicants for jobs -> avoid (implicit) bias

• = Avoiding (negative) emotions following a decision

• i.e. When we make a choice and see how someone with another choice made the better choice, we feel regret -> we want to avoid feeling regret

• The utility of a choice/chosen object is

  • U(A) = v(A2) + r x (v(A2 - v(B1))

    • U(A) = utility of choice

    • v(A2) = Value of what we actually chose (so Option A2 instead of A2)

    • r = regret constance

    • B1 = the value of the other choice which we did not take

    -> (r x (v(A2) - v(B1)) is also called the “disutility function” because it leads to the risk of the value of the choice we made diminishing -> if we exclude/ignore this, all that remains is the value of the choice we made

• This might also explain why people try to gather a lot of information before making a choice but not anymore after making a choice

• Maximizers sometimes still look for additional information after making a decision

• People have a choice between two investments of 1000$

  • 1: government bond, return between 1’000 (no gain) and 1800$ (+800)

    -> risky choice

  • 2: bank account, return between 1250$ and 1300$

    • Gain ensured, but highest gain significantly lower than in the government bond -> safe choice

• Depending on the condition they were in, people made different choices:

  • When they would not be revealed the value of the choice they did NOT make, people chose the safer iotion more often (5:4 ration)

  • When their friend chose the riskier option and they would thus be told what they would have recieved in the riskier option, they chose the riskier option more often (5:3)

• The reveal option might’ve activated their regret aversion, thus leading them to make a riskier choice

• Dutch lotteries:

  • There are draws of random numbers where one person wins or draws of postcodes, where everyone with that postcode who participated wins

  • In the postcode option, regret aversion is likely -> You might learn that your neighbour won and that could’ve been you

    -> regret aversion should increase sales here

• When people have a vivid emotional association of something, they evaluate it as more likely and thus want to avoid it more, even when it logically does not make sense in terms of likelihood

• Examples:

  • People are willing to pay 12$ for an insurance against terrorism vs. 14$ for an insurance for any reason, even though the second option covers terrorism aswell plus many other causes and thus also includes way more likely scenarios

    • BUT: terrorism has a way more vivid emotional association than the very vage “any reason”

    • Terrorism is also broadly overrepresented in the news and thus distorst the memorized frequencies

• Step by step, it looks a bit like this:

  1. Scenario

  2. Memory vailability (availability heuristic)

  3. Vivid emotional association

  4. Results in Risk perception

• Overoptimism: Overestimating the likelihood of good events happening in the future

  • i.e. When people had to predict the likelihood of certain good events happening to them in the next month and this was later compared to what actually happened, people tended to overestimate the good stuff that would happen to them

    • It could be that we process positive information more than negative information

• Depressive realism: idea that people who are more depressive are simply more realistic (which leads to depression, therefore overoptimism has a healthy function)

  • Kind of shown by the data, but depressive people seem to be biased in the other direction -> overpessimism

• Two good reasons for overoptimism

  1. Oversetimating success likelihood seems to make you try harder -> self-fulfilling prophecy

  2. Overoptimism leads to more illusion of control and higher happiness -> better mental health

• Idea, that emotions do influence our decisions, but in a way that is good and necessary

• Step by step:

  1. Enviromental cues (i.e. choice options) trigger somatic markers (i.e. higher heart-beat)

  2. This leads to (the perception of) a certain emotion

  3. This has an adaptive influence on decision-making

• Operationalization

  • People had to solve Wisconsin card sorting task

    • People have to pick cards from four staples

    • In two staples you can win 100$ or lose 100$, in other two you can win 50$ or lose 50$

    • However, the 100$ staple contains a lot more loss cards and the 50$ staple contains a lot more win cards

      -> People have to implicity learn that the 50$ staples are the better choice

  • The participants included healthy participants and patients with bilateral lesions in the vertromedial pre-frontal cortex

    • People with those lesions are believed to have brains which cannot include emotions in the decision-making process

• Results

  • Healthy people stopped taking cards from the 100$ staple after a certain amount of trials, while patients with bilateral lesions in the ventromedial prefrontal cortex kept making their choice randomy

    • Bilateral patients also showd less skin conduction etc., therefore no emotional reactions to their choices

      -> “gut feeling” might be good sometimes, having a feeling for what is good

      -> not including emotions in decision can be bad for the decision making process

• Cognitive evaluation

  • Positive evaluation of performance vs. negative evaluation of performance

• Affective evaluation

  • Positive: satisfaction, surprise, enthusiasm

  • Negative: disappointment, anger, rage

• Together, CE and AE lead to the likelihood of

  • Staying/Switching provider

  • Complaining (thorugh provider or through word of mouth)

• Model to explain satisfaction with a product based on expectation

• We normally do not evaluate products in a vacuum or on an absolute scale, but rather in comparison with what we expected of them

• When it exceeds our expectations, it leads to excitement

• When it is pretty much what we expected, the reaction is neutral

• If it disappoints the expectations, it leads to disappointment

• Brand attachment is beneficial to companies because it

  • helps them predict future growth

  • gives them a competitive advantage in the fight for a limited resource (consumers)

• Ways companies create brand attachment:

  • Increase exit costs: making changing to another company hard/ connecting it with certain costs -> Structurally generated attachment

    • Contractual commitment: getting a cheaper subscription for two years which cannot be canceled

    • Closed eco-systems/special knowledge: when People learn how to work with one product they might be reluctant to swicht and thus having to learn anew how to work with another product

  • Build social relationships, i.e. pretend to be friends with them, attach social identity to brand -> psychologically constructed attachment, i.e. “Identification”

• Market pricing: Rational cost-benefit relationship

  • First level -> transactional

  • Calculation of what you give and what you get and choosing the best option/relationship based on the

  • In marketing:

    • people look for the best cost-value ration and choose accordingly

    • The brands identity is irrelevant

• Equality matching: reciprocal relationships

  • Still kind of transactional but theres more loyality and trust

    • Higher level of emotional/consumer attachment than in market pricing

  • Deal: Company should feel responsible for making customer happy and in turn the costumer stays loyal to the company

    • i.e. You fix the phone I bought with you and I’ll buy my next phone from your company again

• Communal sharing: communal relationship

  • Product and consumer create a social relationship based on (percieved) common goals and values

  • Customer sticks with company and might identify with it but also expects company to follow their shared goals (i.e. ecofriendly production)

  • Some kind of attachment, almost friendship

  • Highest emotional/customer attachment from three levels

• When people believe to have a real relationship with a company, a breach of norms is percieved as a fairness violation

  • distributional fairness

  • procedural fairness

  • interaction fairness

• This leads to a percieved betrayal of trust

• Two possible reactions (love becomes hate hypothesis)

  • Retaliation/revenge

    • Complaint

    • negative review

    • Attempt to hurt the reputation of the company

  • Demand for reparation

    • Through customer services

    • Through third-party consumer protection services

• Law of demand: when prices go up, demand decreases, when prices go down, demand increases

• Veblen goods: when the price goes up, the demand goes up aswell

  • Often luxury goods that increase in value the more difficult it is to obtain them (rare, expensive)

  • Higher entry costs

• Why do Veblen goods exist?

  • Veblen goods as positional goods - psychological motivations:

    • Singal relative status (status symbols)

    • Signaling belongingness to a certain social class and differentiate oneself from other classes

    • signal uniqueness (“individualism”)

    • Standing out (Conspicuous consumption)

• Connected to Veblen goods

  1. Role model/influencer adopts new style to signal uniqueness

  2. Others adopt style aswell -> Object not unique anymore, demand for uniqueness increases again

  3. Rold model adopts new style to signal uniqueness

     -> this process might explain fashion trends

• = indirect cost or benefit to an (uninvolved) third party

  • third party = ohter people, animals, future generations, environment etc.

• i.e. Pollution is an own decision, but it als negatively impacts others

• Standard rational choice only considers own outcomes and does not incorporate social concerns/externalities (neither positive nor neagtive) -> (implicit) assumption of selfish preference

• Idea: people are averse to (decisions that will lead to) inequality

• The extend to which people are averse to different kinds of inequality varies between individuals

• When making a decision, the rational choice theory model of utility is expanded as follows to incorporate inequality aversion

  • Disadvantageous inequality:

    U(x) = xi - beta x max(xi - xj, 0)

  • Advantageous inequality

    U(x) = xi - alpha x (max (xj - xi, 0))

    • U(x): utility of choosing/obtaining x

    • xi = value of choosing/owning x for oneself

    • xj = value of me choosing x for someone else

    • alpha = influence weight of advantageous intequality

      -> how much person is bothered by recieving more than other person -> assumption: people dislike decisions that make them better off than others (might lead to feelings of guilt)

    • beta = influence weight of disadvantageous intequality

      • -> how much person is bothered by recieving less than other person assumption: people dislike decisions that make them worse off than others (might lead to feelings of envy or anger)

• Advantageous inequality aversion:

  • The dictator game: people are given a resource (i.e. 10 Euro and are asked to distribute it between them and someone else as they please)

  • If there wasn’t advantageous inequality aversion, all people would keep all the money

  • However, only 36% or participants keep the whole amount of money

    • 13% even gives away more than half

    • 16% does a 50/50 split

• Disadvantageous inequality aversion

  • The ultimatum game

    • Proposer gets a resource which they can divide between themselves and reciever

    • Reciever can then decide to either…

      • Accept the offer, at which point the reciever and the proposer will split the money according to how the proposer proposed

      • Reject the offer, at which point neither gets anything

    • Rational choice prediction: reciever accepts any offer above 0, because it is still a gain

    • Disadvantageous inequality aversion prediction: responders reject some x depending on their alpha

      -> what is seen is people rejecting offers if the proposer distributes the goods very unequally, with the tendency to reject increasing with more trials

      • 5 - 10% acceptance when offered money only 0-10% of total

      • 20 - 25% acceptance when offered money only 10-20% of total

      • 33 - 40% acceptance when offered money only 20-30% of total

      • 55% acceptance 30-40%

      • 80-85% acceptance when 40-50%

      • 95% acceptance offer >/= 50%

  • Also a lot of variance of acceptance across countries and cultures

• Ultimatum game = points to the existence of disadvantageous inequality aversion

• People would rather punish a person who gave them too little from a share even though it leads to them losing money (disadvantageous inequality aversion) than get money at the cost of not punishing disadvantageous inequality -> the role of emotions

  • Emotions are a commitment device -> they lead people to commit to certain actions

  • Emotions are thus also an enforcement device of fairness

  1. Responder recieves low offer

  2. Responder feels anger

  3. Anger leads responder to reject offer

• BUT: emotions fade with time

  • Hypothesis: people accept more unfair offers when time passes as they are not so angry anymore

  • Found to be true -> acceptance rates with delay (d) vs. without delay (wd)

    • 10% offer: 60% acceptance d, 0% acceptance wd

    • 20% offer: 80% acceptance d, 15% acceptance wd

      etc. ->

  • People also accept more advantageous inequality more as time passes!

• People have different “norms/goals” which can be activated (by the environment, cues etc.) which inform how we will act in different situations

  • Normative goal:

    • morality/pareto

    • Goal: act apropriately

    • long term

    • Example: not crossing the street when its red when children are around

  • Gain goal:

    • “rational choice”

    • Goad: guard and improve own resources

    • Medium term

    • Cost-benefit calculation

      • Between caring about others and caring about self, i.e. for example not comitting a crime when the police is there -> focused on future consequences and not just on hedonic needs, but not necessarily about other people’s wellbeing

    • example: not crossing the street when its red when the police is around

  • Hedonic goal:

    • mood driven

    • Goal: feel better right now -> short term

    • No cost-benefit calculation

    • example: crossing the street when its red because one does not feel like waiting

      
      

• Derived from goal framing theory (norm activation)

• Idea that seeing vandalism/urban decay leads to more crime by activating hedonic norms

  • Also called reversal of causlity, because normally the assumption is that more crime in a neighbourhood leads to more vandalism and decay

• Proof for theory:

  • Throwing more flyers on the floor in alley with graffiti sign and graffiti (neglection of rule) than in alley with sign but no graffiti

  • More money stolen from envelope when there was graffiti (27% vs. 13%) or littering (25%)

  • Letter on the floor less likely to be put in letterbox when theres trash on the floor

• Hypothesis: identifiable victims are more likely helped than statistical victims

• Empirical evidence: People were more likely to donate and expereinced more positive arousal when they saw a photo or a photo with a name of the person they were going to donate money to than when they only saw a name or if there was no information of who the money was going to help

• BUT: small overall identifiable victim effect according to meta-analysis (r = 0.05)

• Single victims > groups of victims

• Children

• Victims with picutres vs. victims without pictures

• Victims which could not be blamed for their hardship

• Vividness: emotional stories increase

  • identifiability

  • relatedness

  • empathy

• Certainity effect

  • Statistical deaths are probabilistic while identifiable victims can convey the impression that bad will happen for sure if person does not act

• Stronger norms of helping

  • Not helping a single person might create stronger feelings of guilt and more social blame compared to not solving an abstract problem that affects many people (and should be solved by many)

• Doing good/altruistic stuff for selfish reasons is evaluated more negatively

• Doing something good for selfish reasons is percieved as tainted

• Examples:

  • Palotta teamworks:

    • Charity organization which raised 305 Millions

    • When people found out that the CED earned 400k a year that lead to public criticism and the company collapsed -> loss in charitable giving

  • When people have to judge two vignettes in which…

    • A guy starts working in a coffeeshop where his crush works and only does his job well to impress her vs.

    • A guy starts working in a homeless shelter where his crush works and only does his job well to impress her

      • they judge both actions as equally beneficial to society (altough one is giving out coffee and the other is helping the homeless -> helping the homeless devalued due to tainted altruism) and the altruism person as more immoral

        -> i.e. less moral to do something good for selfish reasons than to do something neutral for selfish reasons

  • GAP: people had to judge GAP in three different conditions

    • Control -> GAP simply making money for profit

    • Altruism -> GAP donating 50% of a certain product to chairty

    • Tainted altruism -> GAP donating, but the donating campaign earning them extra money

    • The tainted altruism led to people

      • disliking GAP the most

      • Judging GAP the least moral and the most manipulative

        -> GAP in tainted altruism condition judged even worse than in the condition where GAP did not donate anything!

• Ways to work “against” tainted altruism bias:

  • If people were presented the coffeshop and the shelter vignette simultanously, the effect disappeared -> people realized their judgement inconsistency

  • Counterfactual: If in the GAP example people were made aware of the fact that GAP could also simply donate nothing, this reduced the tainted altruism effect

• People may automatically engage in counterfactuals for pro-social action but not for selfish actions

  • i.e. If they see someone doing something altruistic they’re more likely of the surplus altruistic action the person could do than if someone is not engaging in altruistic behaviour to begin with

• People are suspicious about the intentions and credibility of social actions for profit

• What hypocrisy is

  • Double standards in pro-social choice

  • Telling people one thing but then doing another

• How people react to hypocrisy

  • If a person condemns an action and then does that action, they are evaluated worse than if they

    • Do not condemn action but commit an action

    • lie: If they lie about commiting an action. but then actually do the action

      -> Hypocrisy is seen as worse than deception/lying

• Why people might react negatively to hypocrisy

  • Condemnation of immoral act signals that one will behave morally in the future -> has reputation benefits

    • Characters are evaluated equally well if they condemn an immoral act as if we get confirmation that they do not do the immoral act

    • Condemning an immoral act leads to a better evaluation than not doing so

  • Condemning something but then doing it anyways is seen as

    • deception

    • Freeriding on the benefits of moral condemnation/behaving morally

• What voluntary ignorance is:

  • (actively) ignoring (learning about) the externalities/social consequences of ones own actions to avoid

    • feeling bad about it

    • Feeling responsible for it

    • damaging one’s positive self-image

• Examples of voluntary ignorance -> simplified version of dictator game

  • Round 1:

    • Dictator has to choose between two options

      • Option A: Dictator gets 6$ - player 2 get 1$

        • rational choice

      • Option B: Dictator gets 5$ - player 2 gets 5$

        • inequity aversion choice

    • 74% of participants choose option B

      • Evidence for social preferences/inequality aversion

  • Round 2: ignorance treatment

    • Each round, theres a…

      • …50% chance that Option 1 (conflicting interests) will happen, wherein…

        • Option A: Dictator gets 6$, player 2 gets 1$

        • Option B: Dictator gets 5$, player 2 gets 5$

      • …50% chance that Option 2 (aligning interests) will happen, wherein…

        • Option A: Dictator gets 6$, player 2 gets 5$

        • Option B: Dictator gets 5$, player 2 gets 1$

    • The Dictators can decide in each round whether they want to know which option (1 or 2) it is in every respective round or not before making their decision

      • Depending on the revelation, dictator might have to make the “social choice” and lose a dollar

      -> 45% of participants decide to remain ignorant, and 85% of those who remain ignorant choose option A (where they make more money)

• What motivated cognition is

  • Misremembering selfish behaviour, possibly to uphold own positive self image

• Examples thereof

  • People were asked to remember how much they donated in the dictator game with incentives (i.e. they would make extra money if they remembered correctly)

  • People were more likely to misremember how much they gave if they were…

    • Stingy/selfish vs. generous

      • Selfish people were more likely to remember giving more than they did while generous people remembered correctly how much they gave

    • Violators vs. Upholder

      • Violator: people who do/did not act in accordance with their own norms, i.e. who gave less than they personally believed would be morally acceptable to give -> misremebred their behaviour to be more pro-social

• Moral licensing: Doing something immoral after doing something moral because you feel like you can allow yourself to act immoral after acting so morally well

• i.e. eating chocolate after having had a salad for dinner

• Counterpart: moral cleansing

• Doing something moral after doing something immoral to “make up” for the immoral act

• I.e. Eating a salad for dinner after having had a pizza for lunch

• Coutnerpart: Moral licensing

• The trust game

  • Two player

    • Trustor:

      • Gets a certain amount of money, i.e 10$

      • Decides how much to give to trustee, knowing that it will be trippled and that trustee can give back the amount they wish to the trustor

    • Trustee:

      • Recieves money from trustor x3

      • Decided how much to give back to trustor

      -> Game based on trust

    • Rational choice of the trustor would be to keep all of it, since

      • there is no guarantee it will come back to the trustor

      • Assuming the trustee is also rational, they will keep all the money for themselves and not give anything back to the trustor

    • Reality: evidence for social preference

      • Most people send half the money to the other person and keep half of it to themselves for safety

• Possible reasons for why trustees send back money

  • Guilt aversion

    • “I feel bad if I do not meet the expectations of the trustor”

  • Inequality aversion

    • “I feel bad if the other one has less than me”

• Operationalization to distinguish between those two:

  • Two player play the trust game, but:

    • Trustor believes money will be multiplied by 3 (and trustee knows that this is the trustors knowledge)

    • Trustee is instead one of those three conditions:

      • Multiplied by 2, 4 or 6 (depending on trial)

  • Expectation:

    • If trustee sends it back due to guilt aversion, they will send back the amount the trustor expects, independently of which condition they are in

      • i.e. if the 10$ are doubled (20$) instead of tripled (30$) the trustess will still send back 15$ (as expected by the trustor, thus getting less)

      • i.e. if the 10$ are quadrupled (40$) instead of tripled (30$) the trustess will still send back 15$ (as expected by the trustor, thus getting more than the trustor)

    • If trustee sends it back due to inqueality aversion, they will send back half of it, independently of the trustors expectations

      • i.e. if the 10$ are doubled (20$) instead of tripled (30$) the trustess will still send back 10$ (even though the trustor expects more, leading to both having the same amount)

      • i.e. if the 10$ are quadrupled (20$) instead of tripled (30$) the trustess will send back 20$ (even though the trustor expects less, leading to both having the same amount)

• Key takeaway: Asymmetry in information can lead to low trus and low quality goods

• Lemon markets: distinction between

  • Lemons: cars with low value (which should have low cost)

  • Peaches: Cars which have a high value and a high cost

• Conflicting interestes:

  1. Buyers often lack knowledge and cannot perfectly estimate values

  2. Seller always try to sell all cars (including lemons) as peaches, as that will generate more profit

     -> Buyers know that sellers will do that, therefore they are only willing to pay an intermediate price between lemon and peaches (so they won’t massively overpay for a lemon)

  3. This means that peaches are either not bought (because they are more expensive than what buyers are willing to pay) or underpaid -> Peach owners are not willing to sell their peaches on the used car market since they won’t get the amount of money they desereve

  4. This leads to the car market being overrun by low value lemons

  5. Since buyers keep getting lemons for the intermediate price, they further lower the price they are willing to pay -> spiral, where the quality of the cars in the used market decreases and the price buyers are willing to pay also decreaes

     -> Information asymmetries between sellers and buyers can drive out high quality goods -> the bad drives out the good (adverse selection)

  6. This leads to reduced trust in the marked and demand for institutional intervention (i.e. third-party testers, quality control etc.)

• Definition of betrayal aversion

  • Betrayal hurts and has a cost beyond money, so people try to avoid being betrayed

• Operationalisation of betrayal aversion

  • People have choice between

    • A: 10 points for sure

    • B: Either 15 points (with probability p) or 8 points (with probability 1 - p)

    • People have to say the minimum p probability they would accept to choose B

  • Three conditions

    • Decision Problem -> choice between…

      • No externalities

      • Outcome of B randomly assigned with given probability by computer

    • Risky Dictator Game

      • In condition A, chooser and other person both get 10 points

      • In option B, chooser gets same points, but other player gets additionally 15 points for the 15 points condition and 22 points for the 8 point condition

      • Still randomly chosen by computer

      • However, in option B second player might get more than oneself (inequality aversion, jealousy etc come into play)

    • Trust Game (risk of betrayal)

      • In condition A, chooser and other person both get 10 points

      • In option B, chooser gets same points, but other player gets additionally 15 points for the 15 points condition and 22 points for the 8 point condition

      • Also has externalities

      • BUT: player 2 gets to choose which option they prefer if it gets to option B (but essentially the chooser still choses the probability of that choice, so it should not make a difference in what they end up getting compared with other two situations)

    • Results

      • People willing to accept similiarily high/low p for decision and risky dictator condition

        • 60% p = .3 or less

        • 85% p = .5 or less

        • 100% p = .8 or less

      • People choose significantly higher probability for betrayal condition

        • 25% p = .3 or less

        • 40% p = .5 or less

        • 90% p = .8 or less

        • 100% p = 1.0 or less

      -> rational decision makers shouls choose same p across all conditions

    • Conclusion: People behave as though there is a psychological betrayal cost above and beyond any dollar losses

• Average BA (difference risky dictator game and betrayal condition): .15

• Women slightly more BA than men (.15 vs. 14)

• More BA in USA (.22) and Oman (.25) than in Turkey (.16) and Switzerland (.11)

• Differences not significant in all countries

• General condition:

  • Every participant is assigned a mouse

  • Participants can choose to let the mouse live and get nothing or get a certain amount of money to have the mouse be killed

  • Question: For how much money are people willing to let the mouse die (externality)?

• Three conditions

  • Individual condition:

    • Participants simply say the minimum price for which they would be willing to let the mouse die (max 10$)

  • Bilateral bargaining condition

    • There’s a buyer and a seller (the seller being the participant)

    • There are 10 bargaining rounds

    • Each round, the buyer makes an offer between 0 and 20$ to the seller

      • Seller gets amount of offered money, buyer keeps the rest

    • Seller can accept offer or reject it -> if it’s rejected it moves on to the next round and seller cannot go back to previous offer

  • Multilateral bargaining condition

    • There are 7 buyer and 9 seller (the seller being the participant)

      • 2 sellers will not get any money by default -> competition

    • There are 10 bargaining rounds

    • Buyers make offers simultanously -> if sellers see offer they like, they have to make a deal fast (first come, first serve)

    • Buyers and sellers that make deal leave the market

      -> Social Norm activation

• Results

  • Willingness to kill the mouse:

    • Individual round: 40% accepted to kill the mouse for 10$

    • Bilateral bargaining: 70% accepted to kill the mouse for 10$ or less

    • Multilateral bargaining: 75% accepted to kill the mouse for 10$ or less

  • In the bargaining conditions, the price people were willing to get to kill the mouse decreased

• Interpretation

  • Markets can erode moral values

    • Norm activation does not really seem to play a big role here since there was no big difference between the bi- and the multilateral bargaining conidtion

  • People with moral standards can abstain from trading, but cannot influence the price/market

  • A buyer just has to find “the right seller”

• Under perfect conditions (in markets), costs to third parties (environment, workers etc.) are likely externalized since this creates competitve advantages (lower production costs, lower consumer prices)

• Thus societal interventions are needed

• Putting a price on externalities (i.e. carbon emission) can force companies to pay for it and thus respect externalities more

• Intelligence > morality

• Study which proves that

  • General condition: People fill out a cognitive test (50 points max)

    • Condition 1: no externalities

      • 1a: People are told it is an IQ test

      • 1b: People are told it is a general questionaire

    • Condition 2: For each point a person scores in the test, the probability of a mouse being killed goes up by .9 -> morally optimal choice would be to answer all questions incorrectly

      • 2a: People are told it is an IQ test

      • 2b: People are told it is a general questionaire

  • Results

    • Condition 1: no real difference in scores between 1a and 1b

      • 0% scored < 30 points

      • 40% scored between 30 and 40 points

      • 95% scored between 30 and 45 points

    • Condition 2a

      • 10% scored 0 points

      • 12% scored 30 or less points

    • Condition 2b

      • 10% scored 0 points

      • 30% scored 20 points or less

      • 40% scored 25 points or less

        -> did significantly worse to increase chance of saving mouse

    • Conclusion

      • Wanting to be identified as intelligent reduced moral concerns

        
        

    • 
      

      • 
        

      
      

• Study where they started fining people for picking up their children too late

• Result: People started showing up even later after introduction of fine

  • And kept showing up even later after removal of fine

• Interpretation

  • Change in perception: paying to come late (late pickup as a service/fine as a price)

  • Change of relationship from a communal (without fine) to a transactional relationship

  • Formal punishment can reduce instrinsic norm-following

• Classic decision models assume that decision makers only care about own outcomes

• new decision models aim to incorporate social preferences to predict human choice when externalities are present

• Goal-frame theory

• Broken window

• Identifiable victim effect

• Tainted altruism

• Moral hypocrisy

• Moral licensing/clensing

• Guilt aversion (instead of inequity aversion)

• Motivated misremembering of past actions

• Voluntary ignorance

• In game theory externalities are mutual

  • Consequences are not just positive or negative for other person, but also for self

  • Own decisions affect others

  • Others decisions affect self/own outcomes

• Uses the game analogy -> takes ideas from games into real world

• Prisoner dilemma

  • Example: cleaning the dishes

  • Agreement will lead to lightly bad or lightly good, equal outcome for both

  • Disagreement is best for single player, but will cause most costs for other player

    
    

    	Clean	Dont Clean
    Clean	+ I +	- - I + +
    Dont Clean	++ I --	- I -

• Penny game

  • Example: soccer shooter and goalie

    • There’s no solution which is best for both; disagreement is best for one player, agreement best for the other

      
      

      	Shoot corner A	Shoot corner B
      Jump Corner A	+ I -	- I +
      Jump Corner B	- I +	+ I -

• Battle of the sexes

  • Example: couple wants to do something together but they have different preferences -> worst outcome if they have to do anything on their own

  • Agreement best outcome, but to different degree for each one

    
    

    	Movie	Concert
    Movie	++ I +	- I -
    Concert	- I -	+ I ++

• Chicken game

  • Example: going straight with car or dodging

  • Similiar to prisoner dilemma

  • Disagreement best for one, worst for other

    • One agreement absolute worst for both (while in the prisoner dilemma it is still better than disagreement for one player), one agreement kind of in the middle of worseness scala

      
      

      	Dodge	Straight
      Dodge	0 I 0	- I +
      Straight	+ I -	-- I --

• Part of game theory

• Puts different decisions multiple players can take and their respective outcomes for each player into a matrix

• Assumptions/restrictions

  • Actions are taken simultanously (without knowing choice of other player)

  • Game is played “one shot”/one interaction -> not repetitve interaction

• Part of game theory

• Set of strategies for which no player can unilaterally change her strategy and get a better payoff

• No player has anything to gain by changing own strategy -> will be worse off by changing it

• Can also mean: a state in which all players play the best response to each other’s strategy

• Game where everyone has to choose a number between 0 and 100

• Person closest to 1/2 of the average wins

• Common knowledge of rationality assumption -> assumes that

  • Everybody is rational

  • Everybody knows that everybody else is rational

• Based on this assumption, the best choice should be eventually 1

  -> everything above 50 does not make any sense, since the half of the highest number (100) is 50 (people would have to choose numbers above 100 to get a higher half than 50)

  • Choosing a number above 50 is strictly dominated by (always worse than) choosing 50

• Assuming others know that, it makes sense to choose half of 50 -> anything below 25 -> etc., leads eventually to 1

  -> unique Nash equilibrium

• Two verndors have to spread out on beach with ten people spread out in a line

• Whicheverone is closer to people will be the one the people choose to go to

• This will inevitably lead to the vendors being both positioned directly next to the middle of the ten people line (which is not optimal for buyers, but it is a Nash equilibrium)

• Reason why shops selling same good often aggregate close to eachother (i.e. burger King next to McDonald) or why during the election politicians views tend to move more to the center

• Strategy one: look at best strategy for player 1, best reaction of player 2 to that choice etc. until there is no better option for anyone anymore

• Mixed Nash equilibrium: depends on indifference point of both players

• Calculating of expected value for each action based on probability of other player acting a certani way

• Example: prisoner dilemma

  
  

  	Dodge (q)	Straight (1 -q)
  Dodge (p)	0 I 0	-2 I 2
  Straight (1 - p)	2 I -2	-5 I -5

  • Expected value straight: -2 + 2q

  • EV straight: 2q + 5q - 5

  • -2 + 2q = 2q + 5q - 5

  • 2q = 7q - 3

  • 3 = 5q

  • q = 3/5 -> Player 1 should dodge if theres a chace of less than 60% that player 2 will dodge

• Strict dominance: Choice A strictly dominated by choice B -> choice B always better than choice A

• Weak dominance: Choice A weakly dominated by choice B -> choice A is clearly better if other player chooses certain action, but equally as good/bad as B in other option (but there is no option where B is clearly the better choice)

• Level-k bound rationality model -> people have different rationality levels

  • Level 0 rationality: picks a number at random -> m = 50

  • Level 1 rationality: best response to level 0 -> m = 25

  • Level 2 rationality: best response to level 2 -> m = 12.5

  • Level 3 rationality: best response to level 3 -> m = 6.25

  • etc.

• Allows to empirically test at which level people are and what they expect from others

• Median normally at 17, mean at 27

  • Most people between level 1 and level 2 rationality

• But: Choices move more and more towards smaller numbers with every additional round

  -> people might not be super rational, but they’re still able to learn form experience which will bring them to Nash equilibrium eventually

• choices which benefit another person but come with a cost for oneself

• defection: choosing not to help -> no cost + no benefit to other person

• Cooperation is mutually beneficial

• Exploiting cooperation can be rewarding

• Cooperative relationships are risky

• If two cooperative people meet -> cooperation

• If two deflectors meet -> deflection

• If a cooperative person and a deflective person meet -> deflector takes advantage of cooperative person

  • In a mixed population (of deflectors and cooperators), the EV(evil) > the EV(cooperation)

  • Therefore populations should slowly move towards having more deflectors than cooperators

• Degree to which two beings are related

• Calculated as r = 0.5^n

  • n = number of generational links

  • i.e. for parents - children = 1, grandparents = 2 etc.

• Inclusive fitness = direct fitness + indirect fitness

  • Direct fitness = individual reproductive success across the lifespan

    • Quantifiable as the average contribution to the gene pool in the next generation based on own offspring

  • Indirect fitness = reproductive success based on survival and reproduction of relatives

    • Quantifiable as the average contribution to the gene pool based on relative’s offspring

• Helps explain why cooperation makes sense/why beings cooperate (to enhance indirect fitness)

• Rule of how cooperation develops based on (indirect) fitness

• Cooperation develops if:

  • c < rb

    • c = cost of helping

    • r = degree of relatedness

    • b = reproductive benefit

  • Deciding wether to help based on:

    • rb - c

  • the cost of cooperation (c) needs to be

    smaller than the reproductive benefit of

    player 2 (b) weighted by the relatedness

    of player 1 with player 2 (r)

• Explanation for why people cooperate

• Would make sense if there was a perceptible trait which tells us who is cooperative and who is not (like a green beard) so we can cooperate with those people but not with people without this trait

  • If we meet someone with this trait: mutual cooperation

  • If we meet someone without this trait: mutual defection (even if both would be cooperators)

    -> Consequence: cooperation remains group-bound

• But: Very unlikely, that single gene determines altruism and perceptible trait in humans

  • Coding for percetible trait and altruism does not have to be perfectly correlated, suffices if both is inherited imperfectly

• Cooperators start to group together -> defectors try to take on trait and invade those groups -> new trait for cooperators emerges

  • Leads to cycles of cooperation -> exploitation -> colorchange -> cooperation -> …

• Indirect fitness

  • Assumption: Individuals can identify relatedness/phenotype characteristics

  • Individuals cooperate because of inclusive fitness generally (indirect fitness more specifically)

  • Formul

• Yes, people tend to cooperate more in prisoners dilemma the higher the probability p is that they will have to play another round

• When p is 0, then less than 10% cooperation

• Cognitive requirements

  • indirect fitness

    • Recognition of kin

  • direct reciprocity

    • Memory

  • indirect reciprocity

    • Communication

  • generalized reciprocity

    • Mirror neurons

    • Norms & rules

    • Warm glow

• Social complexity

  • indirect fitness

    • Close kind

  • direct reciprocity

    • Dyads

  • indirect reciprocity

    • Groups

  • generalized reciprocity

    • Societies

  
  

• Definition of the game

  • Certain number of players, for instance n = 4 players, get the same resource (i.e. 20$) and can choose between…

    • Keeping a certain amount/all of it for themselves

    • Adding certain amount to the pot -> this amount will be multiplied by a certain amount and shared among the group members

  • Best response - regardless of what others do - is always to keep own amount (i.e. not to cooperate)

• Problem:

  • Ideal outcome for group: everybody cooperates

  • Ideal outcome for self: free-ride on contribution of others

  • tragedy: cooperators stop cooperating if other defect but not vice versa -> game always ends in people not cooperating -> cooperation declines/public goods are under-provided

• Public goods

  • non-excludable: individuals cannot be (easily) excluded from its consumption

  • non-rivalrous: consumption by one consumer does not prevent simultaneous consmuption by other consumers

  • exampes: air, climate, knowledge

• Private goods

  • Rivalrous and excludable

  • Examples: own shirt, food, etc.

• Club goods

  • Excludable but not rivalrous

  • Cinemas, Sports clubs, public transport

• Common pool resource

  • Rivalrous but not excludable

  • “s het solang s het”

  • Examples: free public transport, fish stock (fishing)

• Cooperation creates benefit b

• Cooperation creates cost b for cooperator

  • Benefit for the cooperator: b-c

  • Benefit for the free-rider: b

    -> how can b > b-c? (problem, because this implies individual interes > group interest)

  • Cooperation has to become more attractive than freeriding -> reduce temptation/benefit to free-ride

    • Different techniques, i.e. punishment

• Two studies

  • 
    

• Experiment 1

  • Experiment where people were matched with different person every time

  • Switch in role: once helper (who could donate) once reciever (who recieved donation

  • People were anonymous, i.e. could not tell if they had interacted with person before

  • BUT people saw other persons score, i.e. how much person had donated in past rounds

  • Result: people with lower score got significantly fewer donations as rounds went on (cannot be due to direct reciprocity since they did not recognize former parterns)

• Experiment 2

  • Before game, people could choose network they wanted to play with (request & acceptance of request), continued as game went on

  • Three conditions:

    • Global network: people simply see other peoples network, i.e. who belongs to which group and joins/leaves

    • Global reputation: people saw for every round how everyone acted (defect or cooperate) even from other group

    • Local reputation: people learned how other people in own network acted in terms of cooperation

  • Results

    • Cooperation level

      • Global network knowledge did not do much

      • global reputation knowledge led to more/more stable cooperation levels

      • local reputation knowledge/lrk + gnK led to less cooperation

        > people might’ve wanted to be seen as cooperative by other groups so they would be invited to those groups

    • Density of network

      • in LRK only and LRK/GNK density decreased drammatically -> not many new connections, number of new connections decreased

      • When global reputation was known, density of networks increased constantly -> more new connections happened

      -> global reputation knowledge helps to…

      • isolate defectos

      • let cooperators make links

      • leads to closely interconnected network of cooperators

  • Drug market study

    • in illegal drug markets, drug dealers who deliver get better reviews and thus can sell their products for higher prices

• Study with public good game with three different conditions

  • Basic: players contribute, get results where they see how everyone else contributed and then move on to next round with new players

  • Gossip: Basic: players contribute, get results where they see how everyone else contributed and can leave notes about specific players which those players will take into next round with new players (i.e. this person did not contribute)

  • Ostracism: Basic: players contribute, get results where they see how everyone else contributed and can leave notes about specific players which those players will take into next round with new players (i.e. this person did not contribute) and new group can vote to exclude player from round

• Results:

  • Least cooperation in basic

  • Less steep decline in cooperation in gossip, but still decline

  • Increase in cooperation in ostracism condition

• Intergeneartional public goods problem: when usage of one resource by current generation influences whether/how it can be used by future generations

  • Direct and indirect reciprocity not possible here

  • Peer punishment/ostracism not possible here

• Study: in unregulated treatment (i.e. everyone can decide how much they take), game barely lasted for 2 generations

• Solution: when group voted on how much each member should take maximally (calculation with median), the generations lived on indefinitely

• Counter-punishment: retaliatory punishment (i.e. defectors punishing coopeators because cooperator punished defector earlier)

• Anti-social punishment: defectors punishing cooperators for cooperating

• Rate of antisocial-punishment and counterpunishment can explain variability in cooperation curve in public goods games among countries

  • Countries which did not reach stabel cooperation tend to be countries which use a lot of anti-social punishment (sometimes as much as social punishment)

  • More antisocial punishment in countries which

    • Value norms of civic cooperation less

    • Have less trus in rules of society (courts, police etc.)

      -> correlations, direction/causation unclear

• Cooperation declines less steeply than without punishment, but still declines

• People who cooperate less tend to fake reputation more -> tend to draw the biggest benefit in the end

• Group identity influences punishment given for lack of cooperation

• Three roles:

  • Reciever (R): recieves what dictator gives them

  • Dictator(D): decides to give/not give to R

  • Punisher (P): decides severity of punishment for dictator

• Variation of who was in which group (in/outgroup), dependent variable severity of punishment

• Severity from lowest to highest

  • D and R same group, P outgroup

    • Punishment only if D gave less than half of what he had

  • P and D in same group, R outgroup

    • Punishment only if D gave less than half of what he had

  • P, D and R all in same group

    • significantly higher punishment

    • Punishment almost equally high if D gave 10% or 40%

  • P and R in same group, D outgroup

    • highest punishment

    • Even some punishment if D kept more than 20-30%

  
  

• how much more people send over in dictator game if trustee belongs to same group -> how much more certain people trus people from same group vs. outrgroup

• varies across countries