Cooperation and Coordination - Hardman Chapter PDF

Task 8 - Cooperation (and fear, greed, guilt, punishment & trust) Cooperation and coordination - Hardman chapter Game theory and behavioral game theory Game theory - the analysis of interactions between rational agents, each of which has 2 or more strategies ava...

Task 8 - Cooperation (and fear, greed, guilt, punishment & trust) Cooperation and coordination - Hardman chapter Game theory and behavioral game theory Game theory - the analysis of interactions between rational agents, each of which has 2 or more strategies available to them. Strategies are associated with payoffs and depend on the action the other person takes. In reality, people do not demonstrate perfect rationality, but game theory proposes that people will converge towards rational solutions as they gain experience with the game. Game theory assumes that the players share common knowledge of the rules and assume each other's rationality. Ultimatum game - proposer offers a portion of a predetermined amount (e.g. $10) to the responder. If the responder accepts the offer, then the money is distributed as proposed. Otherwise, neither party gets to keep any money. In experimental settings, there is no communication between proposer & responder and they are anonymous to each other. Analytical game theory proposes that the responder should accept any amount that is offered, because even if it is $0.01, he/she will have more than before. However, people actually behave in a different way than that suggested by this analysis. The empirical results are highly consistent: the offered money is 40-50% on average, and these are normally accepted. Offers below 20% are rejected about half the time. Behavioral/psychological game theory - the study of how people actually behave in interactive situations. Cooperation The prisoner's dilemma and public goods games I skip the definition of prisoner's dilemma, we've had it plenty of times. Because it assumes that both agents are rational, game theory will predict that in the prisoner's dilemma both parties will defect. If A thinks that B will defect, it is more beneficial for A to defect. Also, if A thinks that B will not defect, it is more beneficial for A to defect once again. Nash equilibrium - a combination of strategies that cannot be improved upon, as long as the other player sticks with their chosen strategy (some games have such an equilibrium). In the prisoner's dilemma, joint defection is the Nash equilibrium. In reality, people cooperate more than game theory predicts (1/3 of the time, more on repeated plays of the game). Public goods game (a.k.a. resource allocation task) - individuals are provided with endowment and then given the option to contribute some, or all, or none of it to a pool of resources. After a round during which contributions may have been made, whatever is in the pool of resources is increased by some proportion. Then the resources are divided up among all participants. The collectively rational action is to contribute to the pool whereas the individually rational action is to withhold contribution (and benefit from being a free-rider). The task can also take the form of withdrawing resources from a limited pool, thus risking the depletion of a public good. The evolution of cooperation Computer simulations show that the tit for tat strategy (when considering repeated games, which resemble many real-life interactions) is the most likely to survive evolution. Consideration of others Cooperation in the prisoner's dilemma cannot be attributed to a moral imperative. Study: participants play a series of 1-shot prisoner dilemma games each against different people. Sometimes they are told their opponent's strategy before making their own choice. 37% of games resulted in cooperation when the other person's strategy was not known. 16% / 3% of responses were cooperative when it was known that the opponent had cooperated/defected. Cooperation in the prisoner's dilemma cannot be improved with perspective taking. It is even possible for the opposite to occur. Study: taking the perspective of other groups before a resource-allocation negotiation task led people to generate egoistic theories of others' likely behavior, thus leading them to behave more egoistically themselves. Study: during a prisoner's dilemma tasks, the rate of defection for people asked to empathize with the other person was the same as the control participants (who did the task normally) - 40% vs 41%. However, there was a 68% defection rate among those who had taken the cognitive perspective of the other person. Cooperation in games is not more likely to occur when people think harder about a task. Study: people are asked to verbalize their thoughts while withdrawing resources from a common pool. People in a high cognitive load condition had to hold in mind an 8-digit number while they did the task. Complex Cognition Page 1  had to hold in mind an 8-digit number while they did the task. People in the cognitive load condition were less likely to verbalize task-relevant thoughts and were more likely to withdraw an equal share from the pool. People without the extra cognitive load referred to the task more often and tended to take more than their fair share. Fear, greed and punishment Dictator game - the same as the ultimatum game, but the responder cannot reject the offer. In the dictator game, Cooperation-reducing fear comes in 2 forms: fear of punishment for acting in a self-interested way, or fear for being a sucker (cooperating when nobody else does). Fear partly explains the generosity of most proposers in the anonymous ultimatum game. In the dictator game (the same as the ultimatum game, but the responder cannot reject the offer), proposers offer more, but still more than if they were behaving out of pure self-interest. However, dictators are even less generous when they are convinced that their responses are anonymous to the recipient and the researcher => people often behave in a greedy way when fear is removed. Greed has been found to have more effect on cooperative behavior than fear in prisoner's dilemma. There is evidence of a revenge motive in cooperating people when others defect. Study: above-average contributors to an anonymous public goods game were willing to punish below-average contributors, even though this was a cost to themselves (altruistic punishment). Even though the punishment benefitted the group in the end and can be seen as altruistic, later it was found that the punishers were motivated by anger => the group benefit may have been an incidental effect of revenge. Study: people observe a fair or defecting partner in a previous economic game who experiences pain. If the partner was fair before, activation in pain-related areas was observed (fronto-insular & anterior cingulate cortex). If the partner was not fair, this activation was slightly reduced for women and not significant at all for men. Men also showed increased activity in the nucleus accumbens (associated with reward processing). Nucleus accumbens activity correlated with desire for revenge. Differences in values People can be classified according to their social value orientation. Prosocial people - wish to maximize joint gain and equality in outcomes. Individualists - interested in maximizing their own gains only. Competitors - wish to maximize relative gain (difference between one's and the other's outcome). Study: participants are interrupted halfway through a repeated prisoner's dilemma game and asked to think about the choices they had made and the outcomes that they had received. They were also asked to list some alternatives that were better or worse than what had actually happened More upward counterfactual thinking (UCT) ('It could have been better') was associated with more cooperation in the second half of the game. More downward counterfactual thinking (DCT) ('It could have been worse') was associated with less cooperation. Prosocial people generated the most UCT and the least DCT. Competitors showed the opposite pattern. Individualists generated equal UCTs & DCTs. Therefore, social value orientation was linked to further cooperation, but the relationship was entirely due to the difference in generated thoughts. These findings may not be generalizable to real-life problems. People with more prosocial orientations do not perceive cars as having more negative environmental impact and do not have more preference for public transport. People who engaged in more thought about negative future consequences tended to perceive a greater public impact of cars and preferred public transport (contradicting the results in the study above). Prosocial individuals tend to have more secure attachment styles. Prosocial behaviors also increase with age. Economics majors behave more in accordance with self-interest than non-economic majors in real life and the lab. They are more likely to free-ride in public goods tasks, to offer less and be willing to accept less in the ultimatum game, to give less to private charities, and to defect more often on the prisoner's dilemma. This may be partly due to the training economic students receive. Economic students followed courses in introductory microeconomics. If the course emphasized game theory, the students showed less honesty and more cynicism about others at the end of the semester => certain values can be learned. Culture, cooperation, and economic success Study: 15 small-scale societies around the world play the ultimatum game. Cooperation was stronger in societies where there was more cooperation with non-immediate kin (e.g. in hunting for whales). There was more cooperation in societies where there was greater trading via markets. In 2 societies, people occasionally made offers above 50% but they were often rejected. The authors explained this effect by noting that in those cultures, acceptance of gifts creates an obligation to return the favor, and that favor may be asked at an inconvenient time for the recipient. For this reason, large gifts may be refused. Trust, generosity and communication Cooperation increases when trust in one's partner or others is increased. Complex Cognition Page 2 Cooperation increases when trust in one's partner or others is increased. Trust game - two players, A and B, each have an initial endowment. First, A decides whether to keep his endowment or to send it to B. Then B observes A’s action and decides whether to keep the amount she received or share some if it with A. The experimenter doubles or triples A’s transfer, so that both players are better off collectively if A transfers money and B sends back a sufficient amount. Example trust game - an investor is given $10 and is told he/she could keep it or invest as much as he/she liked. If the investor wants to invest some money, then that will be placed in the hands of an anonymous Trustee in a separate room. The invested money would earn interest and the Trustee will decide how much to give back to the investor. The amount returned can be considered a measure of trustworthiness. Trust and trustworthiness vary considerably by country. People are more likely to trust others who have developed a good reputation. In repeated prisoner's dilemma games, people cooperate much more when playing against the same person than when playing with a different person each time. Study: participants in a trust game were provided with positive, negative or neutral information hinting at their partner's moral character. fMRI was used to measure brain activity during decision and outcome phases on each trial. Although participants are told that their partner's responses may or may not be consistent with their description, 'good' partners were rated as more trustworthy prior to the game and more money was shared with them. After the game, the good partners were no longer rated as any more trustworthy than the bad or neutral partners. The caudate (associated with learning) showed activity in response to partner's choices but not while the participants were deciding. For neutral partners, caudate activity differed depending on whether their decision was to share or keep. For non-neutral partners, there was little caudate activity difference as a result of outcome decisions => prior moral perceptions reduced the extent to which learning occurred as a result of feedback. Activity in the cingulate cortex (associated with conflict monitoring) was observed when the participants decided not to pass money to a good partner, or to share money with a bad partner. The tit-for-tat (TFT) strategy has a disadvantage that when there is noise (erroneous response, e.g. when the situation leads to not cooperating, rather than defecting intentionally), partners can enter a cycle of never-ending retaliation. TFT + 1 - a forgiving tit-for-tat strategy that behaves slightly more cooperatively than the partner did on the previous trial. Study: Using real participants, TFT+1 leads to even more cooperation than TFT. The intentions of TFT were judged as less benign under noise conditions, whereas partners using TFT+1 were judged as equally benign in noise & no-noise conditions. However, sometimes generosity is impossible (e.g. if it costs resources that are not available). In such cases, communication can overcome the problem of noise. Study: participants play a game in which they pass some coins (out of an endowment of 10) to a partner, to whom the coins were worth more in value. The partner was a computer program playing TFT or TFT+1 (passing 1 more coin than the participant previously passed across). On a few trials, noise was introduced by giving the participant fewer coins than the computer had 'intended'. Some participants also received an occasional communication (e.g. 'I wanted to give you 6 coins, but the computer changed my decision') Communication eliminated the otherwise detrimental effects of noise on cooperation, and led people to view their partner just as benign as in a no-noise condition. Study: communication had a bigger effect on people who scored low on dispositional trust. Noise reduced cooperation among those low in trust, but not among those high in trust. When noise was accompanied by communication, there was just as much cooperation in low trust participants as in high trust ones. Emotions can also influence trust in an individual, even though that individual was not the cause of the emotion. Study: emotions were elicited in participants and then they are asked to make trust judgements about an unfamiliar co-worker or acquaintance. Anger led to the lower levels of trust, while gratitude caused the highest levels of trust. Pride and guilt (emotions characterized by personal control) had little effect on judgements. The effect of emotions on trust appeared only for unfamiliar people. Cooperation can increase in social dilemmas when group members are allowed to communicate. 2 explanations are empirically supported: Communication promotes group solidarity. Communication allows group members to express their commitment to mutually cooperate. Greater cooperation occurs when people are allowed to talk face-to-face, because nonverbal cues can be used to communicate cooperative intent. Communication can also decrease cooperation if people do not live up to their promises. Culture and fairness - Heine chapter section The way people judge things to be fair vary considerably across cultures. Distributing resources Principle of need - resources are directed toward those who need them the most. Manifested in most modern industrialized societies through institutions such as universal health insurance or welfare systems. Also governs people's beliefs in distributing resources in many less formal situations, such as contributing to charity or giving money to beggars. Principle of equality - resources should be shared equally among the members of a group. Principle of equity - resources are distributed based on an individual's contributions. Complex Cognition Page 3 Principle of equity - resources are distributed based on an individual's contributions. Meritocracy - a social system that rewards individuals on the basis of the equity principle. Meritocracies tend to be more common in Western & individualistic societies. Meritocracies motivate workers to work hard, which increases productivity, but it can also breed competition among workers, which may disrupt harmonious relationships among them. In the non-Western world and in Western industries that have a more collectivistic base (e.g. through the power of labor unions), the principle of equality is adhered to more. Equality tends to be preserved within a certain range of constraints (e.g. age or official credentials). Seniority system - a system in which time with the company or age is being rewarded. In a seniority system there is no competition between individuals for compensation. Even though theoretically such a system should reduce motivation to work hard, Japanese workers, who are paid more often on a seniority basis, are among the hardest working in the world => the relationship between efforts and the principle of equity/equality depends considerably on the cultural context. Study: in deciding how to allocate bonus money to 2 employees (give most to a needy one, give most to the best-performing one or split evenly), the most popular principle among Americans was to use the principle of equity. In contrast, this was the least popular principle among Indians. A similar study with Australians and Japanese found that Australians preferred an equitable distribution of rewards more than Japanese. Economic games and fairness The first findings (with WEIRD participants) on the dictator game showed that they played the game in economically "nonrational" ways, giving away money that would not provide them with any benefits. The proposed explanation was that people have an internalized motivation for fairness. This was expanded to develop economic theories that large-scale societies and markets were possible because of these internalized fairness motivations. Later multicultural research showed that the offers in the dictator's game differed across cultures and the American sample's offers were at the tail end of the distribution (see figure). This cultural variability could be explained by 2 main factors: market integration and religion. Market integration was operationalized as the % of calories consumed in a society that are purchased in a market. The more experience people had in dealing with markets, the stronger their motivations for fairness. This is the opposite to the initial explanation - markets teach people fairness norms rather than fairness leading to emergence of markets. Religion was operationalized as the % of people who participated in a world religion. Participating in these religions was associated with being more fair at an anonymous dictator game, presumably because of a belief that one's behavior is not anonymous in God's eyes. In public goods games, when people were allowed to punish others, an altruistic punishment (discussed in other text above) emerged. This led to the embracement of the explanation that punishment let to norms for cooperation. However, multicultural research showed that in several societies (e.g. Russia, Saudi Arabia, Greece, and Oman), people showed strong tendencies for antisocial punishment - paying money to punish another player, even if that player was cooperating. The motivations for antisocial punishments were largely a function of revenge, in which people would punish a player who had punished them on a previous round. Even though the tendency for antisocial punishment was negligible in WEIRD societies, for some societies it completely negated the cooperation-enhancing effect. Therefore, the idea that punishment promotes cooperation holds true only in societies characterized by norms for civic cooperation and high levels of trust. Complex Cognition Page 4  cooperation and high levels of trust. Some researchers have proposed that antisocial punishment can be adaptive when the rule of law is insufficient: it's a strategy for establishing status and obtaining the associated benefits when the system cannot be relied upon to deliver those. When people deal with a system that cannot always be trusted to function, they develop alternative ways to ensure their needs are met. In Brazil, there is a complex sociocultural strategy termed jetinho, by which people manipulate or dodge the official rules to achieve things. In societies with higher levels of civic cooperation there is a positive correlation between people’s sense of civic virtue and their happiness. In contrast, in societies with lower levels of civic cooperation there is no correlation between civic virtue and happiness. There may be costs of being exploited by acting virtuous in contexts where trust and cooperation are weaker. After the collapse of the Soviet Union, Russians became suspicious of those who were trying to profit off of the hard work of others. This can be seen in the game Who wants to be a millionaire? - when the 'ask the public' option is used, people tend to choose the wrong answer in order to mislead the contestants. Social neuroeconomics: the neural circuitry of social preferences - Fehr Article Introduction Traditional economic analyses assume that people are exclusively self-regarding. However, people exhibit social preferences - their choices are based on a positive/negative concern for the welfare of others, and on what others believe about them. Social neuroeconomics attempts to understand the brain processes that govern these regular deviations from purely self-interested behavior. There are 2 viewpoints in economic & biological sciences about why prosocial behaviors occur. Behavior in 1-shot anonymous games indicates a reflexive behavior that is highly adapted for repeated interactions in which immediate prosocial behavior earns future benefits. Prosocial behavior reflects robust social preferences for treating others generously/reciprocally. Theories of social preferences and the brain Reciprocal fairness theories - players positively value kind intentions, and negatively value hostile intentions of other players. If player A reduces B's payoff to his own benefit, B will punish A. If bad luck leads to redistribution of income from B to A, B will not punish A. Inequity aversion theories - players motivated by inequity aversion (a dislike of unequal outcomes per se), will also punish other players if bad luck redistributes income. Other theories claim that individuals want to increase the economic welfare of the group they belong to, in order to: a) Experience a warm glow (emotional reward)from altruistic giving to worthy causes (warm-glow giving theory) b) Maintain a positive social image Theories of social preferences are based on the concept of decision utility - a numerical measure that is thought to underlie observed behavior (e.g. the chosen action from a set of choices is inferred to have the highest decision utility). Decision utility can be distinguished from: Experienced utility - the hedonic experience associated with the consumption of a good or event. Anticipated utility - the anticipation of experienced utility at time of decision-making. Social preferences and reward circuitry Theories of reciprocity and inequity aversion imply that in prisoner's dilemma, subjects prefer the mutual cooperation outcome over unilateral defection, even though the latter leads to a high economic payoff. A plausible mechanism would be that people derive higher hedonic value from mutual cooperation, beyond the monetary value. The ventral striatum (activated by anticipated reward) is activated when subjects experience mutual cooperation with a human partner compared with mutual cooperation with a computer partner. Despite identical monetary gain, mutual cooperation with a human partner is associated with higher striatal activity. Social preference theories suggest that individuals tend to punish unfair behavior, like defection in public good and PD games, because not punishing such actions is more unpleasant than incurring the cost of punishing them. A plausible mechanism would be that punishing involves higher activation of reward circuitry. Study: the dorsal striatum (caudate nucleus) is strongly activated in the contrast between a real punishment condition (which hurts the defector) and a symbolic punishment condition (which does not reduce the defector's economic payoff). Study: subjects played a sequential PD game with (confederate) fair and unfair opponents. Then, an electric shock was administered either to the subject or to the confederate partner who behaved fairly/unfairly. When the fair partner received pain, both men and women exhibited empathic responses in ACC and anterior insula. When the unfair partner received pain, only men report a higher desire for revenge, and also exhibit activation in the nucleus accumbens and OFC (this study was also mentioned above). Study: subjects who made more costly donations had more activity in the striatum => giving has its own reward. Study: subjects are in 2 conditions - forced donation (they passively observe that money is transferred to their account or to the charities account) and voluntary donation (subjects could decide whether to accept such transfers). In both conditions, subjects reported higher satisfaction if they themselves receive more money or if the charity receives more money. In both conditions, striatum activations are positively correlated with the money that goes to the charity and to the subjects themselves. Complex Cognition Page 5 Study: fairness of offer (controlled for the absolute size of the monetary game) in the ultimatum game is associated with activations in the ventral striatum. The same offer of $5 elicits higher striatal activation if the offer represents a fair share (e.g. 50%) of the amount, compared to when it represents a small share (e.g. 15%). In sum, social rewards commonly activate the dorsal or ventral striatum. These brain areas that light up when we receive social rewards are quite similar to the ones that activate when we anticipate money rewards or learn from them. This suggests that our brain processes social preferences much like it processes our own monetary rewards when it comes to neural activation. Do activations in reward circuitry predict choices? Some studies suggest monetary gains and losses are both fully processed by a unitary system, centered on the striatum. Therefore, when studies show higher brain activity in certain areas for costly altruistic acts, it might mean that these acts are demanding rather than rewarding. Individual differences in caudate nucleus activation when punishment is costless for the punisher can be used to predict how much individuals actually pay for punishment when it is not costless. Likewise, individual differences in striatal activity in the condition in which donations are forced can be used to predict subjects’ willingness to donate money to charities in the condition in which donations are voluntary. These findings further support the idea that social preferences are rewarding, which in turn suggests that there's a common neural basis for valuing social and other types of rewards The role of the PFC in decisions involving social preferences There is evidence that the PFC resolves the conflict between economic self-interest and social motives. The ventromedial PFC (involved in the integration of separate benefits & costs in the pursuit of behavioral goals) is implicated in the contrast between costly punishment condition and costless punishment of players who behaved unfairly. Subjects with VMPFC lesions reject ultimatum game offers more often, suggesting that the cost of rejecting positive offers has less weight in the decision process if VMPFC is impaired. In charitable donations, the contrast between altruistic decisions involving costs and no costs also activated the VMPFC and the dorsal ACC (which plays a role in conflict monitoring => activity in it is consistent with the existence of a tradeoff between self-interest and prosocial motives). The VMPFC's role in expression of social preferences may be part of a general role in integrating emotional feelings about costs & benefits (economic or not). VMPFC is involved in emotional processing, moral judgment, and integrating the value of consumer products and their prices. VMPFC lesions are associated with poor choices in various situations that require integrating costs and benefits. The DLPFC & VLPFC also play an important role in the processing of decisions involving social preferences. In an ultimatum game, when a low positive offer is rejected (there is a conflict between fairness and economic self-interest), there is higher activity in DLPFC, anterior insula (AI), and ACC, in the contrast between unfair and fair offers. The higher the activation of the AI, the more likely a subject rejects an unfair offer => AI activation may be related to degree of emotional resentment of unfair offers. ACC activation may reflect the motivational conflict between fairness & self-interest when facing unfair offers (ACC has a role in conflict monitoring). DLPFC may represent cognitive control of the emotional impulse to reject unfair offers. Right VLPFC is more activated when unfair offers are accepted, which may indicate that it downregulates the resentment associated with unfair offers. A later study tested the hypothesis that DLPFC represents the cognitive control of the impulse to reject by disrupting DLPFC activity with TMS. TMS of right DLPFC increased the acceptance rate of unfair offers relative to a placebo stimulation (9% to 44%). TMS of left DLPFC did not affect behavior. This suggests that right DLPFC is not involved in controlling the impulse to reject unfair offers but in controlling the impulse that pushes subjects towards accepting unfair offers, (i.e. with controlling economic self-interest). Disruption of the right DLPFC only affects behavior, but not fairness-related judgements. Trust, reputation and social preferences Social preference models suggest that people who care about reciprocity and fairness feel a strong displeasure when trust is broken, in addition to the economic loss. Study: synthetic neuropeptide oxytocin (OT) is infused intranasally to players in a trust game. Players who received oxytocin (OT) were more trusting than those who received a placebo, but their beliefs about repayment and risk- taking remained the same. Oxytocin appears to reduce the fear of betrayal in social situations, which is in line with research in animals that shows it reduces defensive behavior and promotes bonding between individuals. The idea that oxytocin reduces the fear of betrayal by reducing activity in the amygdala is supported by a study that found o xytocin reduces amygdala activity when people look at emotionally arousing images (the amygdala is involved in evaluating the trustwo rthiness of faces and processing uncertain social situations). Trust decisions are also likely to involve perspective taking. Accordingly, they also activate areas implicated in theory of mind tasks (e.g. paracingulate cortex and the posterior superior temporal sulcus). A trustworthy person’s face automatically triggers emotions and reward expectations, as if simply seeing another person’s face activates its representation as a future exchange value. Complex Cognition Page 6 representation as a future exchange value. Study: players face a series of cooperative & noncooperative opponents in a sequential PD game. Simply displaying the faces of cooperative partners (relative to neutral faces) in a subsequent gender-assessment task activated striatal & emotion-related areas, such as amygdala, insula & putamen. The striatum plays a role in learning the opponent's trustworthiness. Study: activity in caudate nucleus signals whether the other player reciprocates an earlier move. The other study is discussed above (CTRL+F for caudate). social rewards activate in social thing brain Complex Cognition Page 7