Behavioural Theories: Behavioural Theories: Behavioural Economics and Organizational Studies Julita Vassileva Social Computing Class 2009 Social Computing Class 2009
Behavioural Economics Behavioural Economics • Based on Based on
How do we make choices? How do we make choices? • “The cycle of relativity”: We are always looking at the things y y y g g around us in relation to others
• Welcome Welcome to the Economist subscription center. Pick the type to the Economist subscription center Pick the type of subscription you want to buy: ‐ Internet only subscription ‐ $59 ‐ Print only subscription ‐ Print only subscription $125 ‐ Print and internet subscription ‐ $130 Do magazines really do this now?
The Fallacy of Supply and Demand The Fallacy of Supply and Demand • We purchase items based on value, quality or availability, sometimes – all three sometimes all three • The methods of appointing a value to an object with no previous value is subject to irrational pricing – E.g. a fancy ad with “equally” precious items – Placing “equally” precious items in a luxury store window.
• Once we buy a product at a certain price, we become y p p , "anchored" to that price • The theory of supply and demand challenged: – demand, the determinant of market prices, can be easily manipulated – supply and demand are dependent on each other (manufacturer's suggested retail prices affect consumers' willingness to pay) – relationships between supply and demand are based on memory rather than on preferences
The Effect of Expectations • Can previous knowledge change an actual sensory experience? • Experiment 1 with two beers: Budweiser and the MIT Brew, (beer containing balsamic vinegar). – If people didn’t know what they were drinking in advance, they liked the MIT Brew – If they were told in advance, they liked Budweiser better. – If they were told after they reported their liking/disliking, they stuck to their choice their choice
• Experiment 2 with female Asian students doing math test • Experiment 3 with organic foods • Higher price sets higher expectations – curious medical studies, generic and brand‐name drugs (the power of placebo) • Outcome: Expectations shape stereotypes – O t E t ti h t t we see what we h t want to see.
The High Price of Ownership The High Price of Ownership • Endowment effect: we overvalue what we have • Inconsistent with standard economic theory which asserts y that a person's willingness to pay for a good should be equal to their willingness to accept compensation to be deprived of the good the good. • Duke team basketball ticket raffle experiment – Those who didn’t win a ticket were willing to pay $170 on average – Those who won a ticket would sell it for $2,400 on average (14 times higher value!).
• Reasons: – ownership is such a big part of our society that we tend to focus on what we may lose rather than on what we may gain. – the connection we feel to the things we own makes it difficult for us to the connection we feel to the things we own makes it difficult for us to dispose of them. – we assume that people will see the transaction through our eyes.
The Cost of FREE! The Cost of FREE! • Humans react to the word FREE! and zero without rationalizing their choices their choices. • Experiment1 : students offered to choose between two chocolates: 1) a 15c Lindt truffle (regular bulk price 30c) and a 1c Hershey and 2) a 14c Lindt truffle and a free Hershey Which was the preferred choice in each case? Experiment 2: A free $10 Amazon certificate and a $20 certificate for seven p $ $ dollars. Which one would you pick (don’t think too much )?
• There is something “magical” about “FREE!” – no chance of loss • Amazon’s “free shipping” deal in France A ’ “f hi i ” d l i F • But what is the real cost of “Free” stuff? – Your Your time? Missing on a better option? Forgetting your priorities? Having time? Missing on a better option? Forgetting your priorities? Having an unwanted item which takes space...
• Use in marketing –Gillette razors , electric toothbrushes …
Procrastination and Self‐Control Procrastination and Self Control • Two states in which people make judgments – Cool Cool state state – rational, long rational long‐term term goals, plans goals plans – Hot state – irrational, seeking immediate pleasure or reward – Procrastination = putting off long‐term goals for immediate gratification
• University experience: – I promise to submit Assignment 1 on ….. (penalty 1% for each day after) – 3 classes: 1) students pick deadlines, 2) no deadlines, 3) prof sets deadlines ) p , ) , )p – Which class did the best?
• With proper motivations such as deadlines and penalties, people are more willing to meet those deadlines or long‐term goals illi t t th d dli l t l • Healthcare experience: – When When patients have to pay a penalty for not showing up for appointment, patients have to pay a penalty for not showing up for appointment, they tend to keep their appointments. – But one has to be careful with penalties – example with a day care
Social Norms: We are happy to do things, but not when we are paid • Social norms govern – friendly reciprocation ‐ requests with instant payback not required – market norms—which account for wages, prices, rents, cost benefits, and repayment being essential.
• Combining the two creates troubling situations, e.g. paying your mother in law for her lovely Thanksgiving dinner – Ask a passerby to help you move a couch, for free and for $2. Ask a passerby to help you move a couch for free and for $2 – Daycare example again
• When social norm and market norm collide – the social norm goes away for a long time. – What is the best way to motivate people to do something? Relying on social norm or market norm ? – It depends on what you are asking them to do – Money is often the most expensive way to motivate people
On happiness On happiness • Dan Dan Gilbert Gilbert’ss TED talk (20 min) TED talk (20 min) http://video.google.com/videoplay?docid=‐ 7822696446273926158#
And for completeness – three slides on classical economics • Prisoner Prisoner'ss dilemma dilemma • Tragedy of commons • Repeated prisoner’s dilemma d i ’ dil
Prisoner’ss dilemma Prisoner dilemma
Prisoner A chooses “defect” (betray the other)
Prisoner B chooses “defect” (betray)
Prisoner B chooses “cooperate “ (not betray)
A – 3 years
A – walks
B ‐ 3 years
B – 6 years
A – 6 years Prisoner A chooses “cooperate” (not betray) B ‐ walks
A – walks
B – walks
Tragedy of commons Tragedy of commons
Hardin (1968) Science Vol 162 Hardin (1968) Science, Vol. 162,
» No. 3859
Economics of cooperation Economics of cooperation • Repeated Repeated interactions interactions – e.g. repeated e g repeated prisoner’s dilemma • “The The shadow of the future shadow of the future” • Strategies – Always Cooperate – sucker outcome – Always Defect – minimizes welfare – Tit for Tat – the optimal strategy for infinite time horizon repeated game Robert Axelrod (1980) The Evolution of Cooperation.
Based on Based on • James Surowiecki, The Wisdom of Crowds James Surowiecki The Wisdom of Crowds • Cass Sunstein, Infotopia
Smart crowds: examples Smart crowds: examples • In In 1920 sociologist Hazel Knight started a 1920 sociologist Hazel Knight started a series of experiments with her students: – Estimate the temperature in the room Estimate the temperature in the room – Estimate the number of beans in a jar – Ranking 10 similar items by size… Ranking 10 similar items by size
• Who wants to be a millionaire? • The Challenger disaster and the stock market
Why? • The Condorcet Jury Theorem: The Condorcet Jury Theorem: – Suppose that voters are answering the same question with two possible answers: True / False question with two possible answers: True / False – Suppose that the probability that each voter will answer correctly exceeds 50%. answer correctly exceeds 50%. – Æ the probability of a correct answer by a majority of the group increases toward 100% as the size of the g p group increases. Corollaries: groups do better than individuals, large groups do better than smaller ones
4 Characteristics of Smart Crowds 4 Characteristics of Smart Crowds • Diversity Diversity of opinion (each person should have of opinion (each person should have some private info) • Independence (people (people’ss opinions are not opinions are not influenced by those of others) • Decentralization D li i (people are able to specialize ( l bl i li and draw on local knowledge) • Aggregation (a mechanism exists for turning private judgments into a collective decision)
Diversity •
Through the history of most technologies the early stages were characterized with a huge number of alternatives characterized with a huge number of alternatives – – – – –
•
Screws Cars, Planes Personal computers p Browsers, Web search engines PDAs / cell phones…
Through a collective search process, some designs win over others – Th The winners are “smarter” (cheaper, more universal, better, prettier…) i “ t ”( h i l b tt tti ) – The search process is more like a biological phenomenon – like ant colony intelligence, or survival of the fittest in evolution, than a deliberative process.
•
Lack of diversity in deliberative processes also leads to failures – The Bay of Pigs invasion in Cuba, 1961…
•
But sometimes it is really hard to find dissenting views – Original minds are precious, in groups peer pressure and authority Original minds are precious in groups peer pressure and authority interfere
Social forces for conformism Social forces for conformism • People can be influenced easily: eop e ca be ue ced eas y: – Solomon Asch’s experiment – “What’s up in the sky?” – Queuing for nothing
• John Maynard Keynes: – “Worldly wisdom teaches that it is better for reputation to fail conventionally than to succeed unconventionally . unconventionally”.
• Cascades result from people valuing more the p public information than their private information p
Information cascades Information cascades • The wooden plank roads (Canada, ca. 1840) cascade swept NY, mid‐Atlantic states, the Midwest. • Mechanism of an information cascade – The The two restaurants model (Bikhchandani, Hirschleifer) two restaurants model (Bikhchandani Hirschleifer) – People’s decisions are not made simultaneously, but in sequence – There is no peer‐pressure, conformity, or mindless trend‐ There is no peer pressure conformity or mindless trend following, they believe that what they choose is better. – William Sellers orchestrated a cascade that lead to the adoption to his screw (became the standard) adoption to his screw (became the standard). – The Telecom bubble in the late 1990ies was a cascade – The house‐market bubble in the US recently….
Starters / breakers of cascades Starters / breakers of cascades • Overconfident Overconfident people people – who value their own who value their own private information more than the public one – People are in general overconfident People are in general overconfident – People are more overconfident when facing difficult problems than when facing easy ones difficult problems than when facing easy ones – Overconfident people are less likely to get sucked into negative information cascades, they go by into negative information cascades, they go by their gut, and disrupt the signal that everyone else is getting.
Good and Bad cascades Good and Bad cascades • Good cascades Good cascades – “Follow the crowd and you will find the buffet” – Imitation of profitable behaviour is usually profitable for the imitator too. – Farmers imitate their neighbors who started growing better crops from new seeds better crops from new seeds.
• Bad cascades: result from people switching off their p private information/reasoning… / g • Yet they are a mechanism for aggregating information, just like a voting system or a market.
Decentralization • Example: Example: – the US security agencies (CIA created in 1947 to centralize all intelligence… but FBI, NSA remained, all formed by independent, self‐interested agents) – Efficient in gathering information, but couldn’t prevent Perl Harbor the bombing of the WTC in prevent Perl Harbor , the bombing of the WTC in 1993, the US Embassy in Kenya, and 9/11. – The politicians The politicians’ reaction has always (1946, 2001) reaction has always (1946 2001) been to call for centralization, opposite to the trends in business, academy, science, technology
What is decentralization? What is decentralization? • Two main features of decentralized systems (DS): – Power does not fully reside in one central location – Many important decisions are made by individuals based on their own specific knowledge rather than by an omniscient or farseeing planner.
• Examples: Examples: flocks of birds, free‐market economies, cities, p2p computer flocks of birds free market economies cities p2p computer networks… and also: the US public school system, corporations • DS foster and are fed by specialization, tacit (context‐based) knowledge • DS allow for independent agents to coordinate their actions and solve DS allow for independent agents to coordinate their actions and solve difficult problems • Problems of decentralized systems: – Coordinating independent, self‐interested agents g p , g – Relevant local information may never get propagated to where it will be useful. – Important to have mechanisms that allow aggregating information.
Coordination • Example: a crowd of pedestrians moving on the sidewalks p p g of a big city • Coordination problems – Definition: problem that to solve it, a person has to think about what other people think the right answer is – They are ubiquitous in real life: when to leave work to avoid y q traffic jams, where to go to eat tonight, how to meet our friends, how do we allocate seats in the subway, who will work where, how much should my factory produce , y yp – Key question: How can people voluntarily (without anyone telling them what to do) make their actions fit together in an efficient and orderly way? efficient and orderly way?
The El Farol problem The El Farol problem • “No one goes there anymore. It is too crowded” ‐ Yogi Berra • Early 1990ies; economist Brian Arthur posted the problem named after a popular bar in Santa Fe – If El Farol is <60% full on any night, everyone there will have a good time If it is >60% no one will have fun Therefore people will go only if time. If it is >60% no one will have fun. Therefore, people will go only if they think it will be <60% full, otherwise they stay @home. – How does one decide what to do an a given Friday night?
• One multi‐agent simulation with a variety of strategies shows wide g y g fluctuations each week, but overall over a number of weeks – average just 60% attendance (optimal solution, max. # people having fun) • Another multi‐agent simulation where agents followed only one Another multi agent simulation where agents followed only one simple strategy (go if your last experiences have been good, don’t worry about the others) showed a gradually stabilizing attendance at jjust under 60%. Æ independent thinking helps! But inequality – p g p q y the population divided itself into two groups: regulars and people who went only rarely, and usually had bad time when they went…
Focal points in coordination Focal points in coordination • Thomas Schelling experiment (1958) Meeting g p ( ) g someone in NY? Where? When? Count suggestions: – The info booth at Grand Central station at noon… The info booth at Grand Central station at noon
• First come – first served seating in public spaces • “Name Name a positive number a positive number” • Focal points – p salient landmarks upon which p people’s expectations would converge (Schelling points) – provided people share a common cultural background cultural background.
Norms and conventions Norms and conventions • How How quickly will the crowd learn which stairs quickly will the crowd learn which stairs to use? • Another Milgram Another Milgram experiment experiment – ask people to ask people to give you their seat in the subway without offering any reason…. • And yet another one – cutting into a long line • Convention and pricing… many companies set Convention and pricing… many companies set their prices based on simple rules and guesswork.
“Fairness”, Fairness , punishment, trust punishment, trust • • • •
The Ultimatum game – as one‐shot, blind game Repeated prisoner's dilemma The public goods game (your assignment 2) Evolution of cooperation, Robert Axelrod, 1980 – Cooperation Cooperation is the result of repeated interactions with the same people. The is the result of repeated interactions with the same people The foundation of cooperation is not really trust, but the durability of the relationship. People who repeatedly deal with each other recognize the benefits of cooperation and do not try to take advantage of each other, because they know that if they do, the other person will be able to punish them. The key to y , p p y cooperation is the “shadow of the future”.
•
The Ultimatum game, revisited – This time, it is played repeatedly between the same two players, iterating their roles… roles…
•
Trade, exchange, cooperation with strangers – The value of trust in trading – evolved with capitalism bringing prosperity for cooperating groups (ethnic groups, hanseatic guilds, Quakers)
Aggregation – weeding out the alternatives l • A decentralized system can only produce intelligent results if there is a means of aggregating the private information of everyone • An aggregation mechanism is a form of centralization, (ideally) of all the private information of the participants – Should provide incentives for revealing truthfully private info – Should not inject extra bias in the system
• Mechanisms: – – – – –
One person with foresight Deliberation Polls / votes Price in a open market Prediction markets Prediction markets
Voting • Simple Simple mechanism with good results if agents mechanism with good results if agents are truthful in their votes • But the mechanism allows for strategic voting But the mechanism allows for strategic voting Æ skewing the results
Markets/Price • Class experiment p – Students assigned roles and max. / min. prices – Market Market – each seller runs an auction, buyers bid each seller runs an auction buyers bid – The prices offered quickly converge to equilibrium
Prediction markets Prediction markets • “Want Want to bet? to bet?” – When people ask for a bet, they are bypassing deliberation and offering a good clue about the deliberation and offering a good clue about the confidence with which they hold their views
• If If we want to get access to many minds, we we want to get access to many minds we look for economic incentives – That is exactly what markets do. Prices emerge as That is exactly what markets do Prices emerge as a product of the independent judgments of many p p people each with their own judgments and tastes. j g
Prediction markets Prediction markets • The central advantage of PM: they give people the g yg p p right incentive to disclose the information they hold (they stand to win money if they get it right, or lose money if they get it wrong) money if they get it wrong) • Those with specialized information can gamble as they see fit • Selection: only those who believe they have some viable information participate (no more empty talkers) • Those that are more confident bet higher stakes • The University of Iowa Election Markets since 1988
Why do prediction markets work? Why do prediction markets work? • Friedrich Hayek, 1945 “The User of Knowledge in Society” • The key economic question: how to incorporate “dispersed bits of incomplete and frequently contradictory knowledge which all the separate individuals possess? knowledge which all the separate individuals possess? • Answer: prices aggregate both the information and the tastes of numerous people, far more material than could possibly be assembled by any central planner or board possibly be assembled by any central planner or board. • Prices act as an astonishingly concise and accurate coordinating and signaling device. They incorporate that di dispersed knowledge and in a sense also publicize it dk l d di l bli i i • When people are willing to put their money where their mouth is, there is an increasing likelihood that they will be , g y right.
Failed predictions: manipulation, bias, b bbl bubbles • Manipulation by powerful speculators • Bias – inherent biases (overconfidence, preferences) in large groups of investors • Bubbles: anticipating the choice of others sets conditions for information cascades Prediction markets can only make predictions on facts, not markets can only make predictions on facts, not • Prediction on values (e.g. deciding on whether to authorize stem cell research can not be done through a market, but a market could potentially predict if a law authorizing stem cell p yp g research will pass). They are not suitable for making decisions, only for predicting what decisions would be….
Deliberation • Group Group deliberation deliberation – the ideal of a the ideal of a “deliberative deliberative democracy”: committees, teams, juries… • Sources of deliberative failure Sources of deliberative failure – Informational influences – Social pressures S i l
• Private benefits versus Social benefits for each member of the group • Example: the Columbia disaster
How are small groups different from wise crowds? d • People think of themselves as members of a team, while in a market, they think of themselves as independent actors. The group has an identity of its own • The group has an identity of its own – Consensus is important for the existence and comfort of the group – Influence of the people in the group on each other Influence of the people in the group on each other’ss judgment judgment is unavoidable. – Group polarization
• Collecti Collective wisdom, in contrast, is something that emerges e isdom in contrast is something that emerges as a result of many different independent judgments, not something that the group should consciously come up with. ith
When are small groups productive? When are small groups productive? • On “eureka” problems – p like crossword puzzles, when p , there is a clear self‐affirming answer which is clearly right once stated (if one member of the group finds it) • When they have a clear agenda and when leaders take Wh th h l d d h l d t k an active role in making sure that everyone gets a chance to speak p • Dissenting opinions should be encouraged, the devil’s advocate role should be played frequently • The diversity of the group is important • The order of speaking is important (early speakers set the tone) the tone)
Measures for successful deliberation • Be aware of status and leadership Be aware of status and leadership – Encourage low status members (women, minorities) – Leader should speak last after listening to everyone Leader should speak last after listening to everyone
• Change reward structures / social norms – Reward group success, not individual; do not discourage dissent / whistle‐blowing
• Pre‐deliberation anonymity, secret ballots, Delphi d lb b ll l h Technique • When silence is golden?
Wikis, Blogs, Open Source: mass deliberation tools? • Wikis Wikis / open source software offer yet / open source software offer yet another way of aggregating knowledge and creativity • Contributors lack economic incentive, but still behave in good faith; wikis and OS won’tt do so behave in good faith; wikis and OS won do so well if the majority of contributors were confused incompetent mischievous biased confused, incompetent, mischievous, biased or destructive • Blogs: ensure widely dispersed and diverse Bl id l di d d di source of information and opinions