Frontiers in Social Dilemmas Research

Social dilemmas are complex situations in which we can choose what is in our own immediate best interest or what is in the best interest of our groups, which include ourselves as well as others. Finding solutions for social dilemmas constitutes one of the most important challenges for the social and behavioral sciences.

A social dilemma is a situation where a collective cost or risk is incurred, taken or generated through the combined negative external effects of various individuals who act (relatively) independently from one another. Vivid present-day examples of collective risk generation via individual activities are: littering of public places by individuals spending some time in them; threats to the quality of public education through individuals’ desires to increase private spending capacity; loss of natural open space through individual preferences for more spacious household premises; over-harvesting of ocean fish stocks for the survival of individual fishing organizations; local and regional air pollution from the use of numerous motor vehicles; and wholesale deforestation of tropical regions for the subsistence of local farmers and cattle-breeders. In many cases, collective risks also increase through the sheer growth in the number of separate actors such as inhabitants, households and commercial enterprises.

The chapter reports results of a panel study on citizens’ attitudes toward sharing and contributing to public goods and services. The study was run in Poland, a nation in transition from a centrally planned to a free market economy, from a totalitarian to democratic political system. A collectivity that is experiencing change of this magnitude will seem quite peculiar unless one understands the psychological context in which the changes are taking place. Hence, the Results Section of the chapter begins with some background information about the current state of mind of the Polish people during the transition. The chapter as a whole, however, has two main purposes. First, we want to describe the attitudes of Poles toward public goods and services and to demonstrate how they depend upon demographic factors. Second, we would like to analyze certain methodological problems associated with measuring attitudes in surveys under conditions of rapid social change.

In an attempt to generalize the GEF hypothesis (H. A. M. Wilke, 1991) to a real-life public-goods dilemma, the main question asked in three studies was whether perceived distributive fairness affects willingness to pay for community child care. In the first study, attitudes towards whether the quality of child care should be distributed equally to all children, according to the needs of the children, or proportional to how much the children’s parents pay were surveyed in 1,840 Swedish parents living in five municipalities of different sizes. Preferences for different methods of payment were also measured. Although the results lent some support to the hypothesis that perceived distributive fairness plays a role, other factors were found to have a stronger effect on willingness to pay. The main survey results were replicated and extended in two additional studies employing a hypothetical society paradigm in which undergraduates were asked to respond to scenarios.

In a recent review, Komorita and Parks (1995) suggest that the formation of coalitions and harvesting behavior in a social dilemma may be considered examples of a mixed-motive interdependency. In contrast, in the present study (see also Mannix, 1991; 1993) on a conceptual level coalition formation and social dilemmas are connected and applied to the problem of dispersion and concentration of resources in organizations. Thereafter, an experimental study is presented investigating negotiations about coalition formation and allocation of resources in social dilemmic organizations, which differ in two respects: the superordinate goal of the organization and the power of participants.

By means of a 2(superordinate goal: concentration, dispersion) × 2(power: symmetry, asymmetry) factorial design negotiation behaviors, formed coalitions and the allocation of outcomes are investigated. In symmetric and asymmetric power organizations, members are equal or unequal in power, respectively. In the dispersion and the concentration conditions over all rounds, the superordinate goal of the organization is best served by large or small coalitions, respectively. In all conditions within a round, each of the group members is best off by becoming a member of a winning coalition, and by obtaining a larger share of the outcomes of a coalition.

The major results of this study in which coalitions are embedded in a social dilemmatic organization are (1) coalitions are formed in agreement with the superordinate goal of organizations, and (2) as for the allocation of resources in equal power organizations Gamson’santi-competitive theory (Gamson, 1964), and for unequal power organizations Komorita’s bargaining theory (Komorita & Chertkoff, 1973) is supported; (3) in unequal power organizations, members expressed to be more competitive and more proposals were necessary to form a coalition than in equal power organizations. The results are discussed in view of the social dilemma and coalition literature.

Two alternative explanations for the effect of social motives on cooperation in social dilemmas were identified. One explanation (a transformation explanation) suggests that those with more cooperative social motives attach greater weight to others’ outcomes. The other explanation (a moral explanation) suggests that those with more cooperative social motives are generally more concerned with acting in a morally-prescribed fashion. The transformation explanation predicts that social motive effects should be moderated by the efficacy of cooperation; the moral explanation predicts no such moderating effect. An experiment that tested these competing predictions is reported. Its results supported a dual-process model—the transformation explanation holds under conditions where it is unclear whether or not cooperation is morally prescribed (e.g., when there had been no prior discussion of the dilemma), whereas the moral explanation holds under conditions where it is clearer that to cooperate is to ‘do the right thing’ (e.g. following group discussion of the dilemma).

In this paper we argue that in social dilemma situations, group members often tacitly coordinate choice behavior on the basis of rules of fairness. Results are presented on the employment of coordination rules in asymmetric dilemmas. Three different research topics are covered, suggesting that tacit coordination of choice behavior is affected by the justification of asymmetries, the presentation of social dilemmas, and environmental uncertainty. On the basis of these research findings, general suggestions are formulated, as well as suggestions for future research on fairness and tacit coordination.

This paper reports two experiments which examined the impact of information about the collectively rational solution on the behaviour of individuals who differed in social value orientation. In the first, Rapoport’s(1988) version of the largest number game1 was given to subjects under minimal, and extended, information conditions. It was predicted that prosocial, or cooperative subjects would show higher levels of cooperation when given extended information, but that pro-self subjects would not, on the grounds that the latter are not motivated to maximize collective utility. Results supported the hypothesis, although absolute levels of cooperation remained low. The second study employed an investment dilemma game, in which subjects started with a fixed amount of money invested, and could withdraw all, part, or none of it. The rules of the game specified that the investment group would collapse if more than half of the fiinds were withdrawn. Cooperators and competitors were assigned to either a basic or extended information condition; in the latter, they were given additional information and examples about which behaviours were necessary to keep the investment account going to ensure that all would gain (joint maximisation), and which behaviours would be most likely to maximise individual gain /avoid loss, if others were not cooperative (individual maximising strategy). Extended information produced significantly higher rates of cooperation only for subjects with prosocial motives, as predicted. Across information conditions, competitors were significantly less cooperative than prosocial subjects. Expectations about others’ investment behaviours were highly correlated with own investments. The pattern of results in the two studies is interpreted as consistent with the argument that information facilitates cooperation only in those who have a prosocial orientation.

When group members share their resources equally, regardless of their contribution to the collective group outcomes, equity norms may be violated. Some individuals may use the opportunity to take a free ride and as a result others will be put in the position of suckers. The equal division rule, however, has a clear advantage that has received little theoretical and empirical attention. Sharing with others reduces the risk associated with individual outcome in an uncertain environment and thus has the effect of insurance. We report an experiment in which we examined subjects’ willingness to share with others as a function of environmental uncertainty and personal risk preferences. Specifically, each subject had to choose between an individual lottery and an equal share of a combined group lottery. We found that, when uncertain about their own outcome, risk-averse individuals prefer the less risky group option, whereas risk-seeking individuals prefer to take their own risk. Assuming that in real-life situations people are typically risk-averse, we conclude that the desire to reduce uncertainty concerning one’s outcomes indeed motivates people to pool their risks with other group members.

An experiment is described in which 61 subjects made choices in nine trials of a give-some game as well as in nine trials of a take some game. Dependent on the trial 10 to 90 percent of the participants had to make a cooperative choice (contributing money in the give some game; not taking money in the take some game) in order to obtain a bonus for all participants. On the basis of the subjects’ choices and their expectations about the choices of other participants subjects’ greed, fear, solidarity and sense of duty were assessed. If a low percentage of the participants had to make a cooperative choice to obtain the bonus, subjects were more cooperative in the give some game than in the take some game while the opposite was the case if a high percentage of the participants had to make a cooperative choice. Whether the level of greed or fear was higher depended on the type of game and the percentage of cooperative choices required.

The focus of this chapter is on cooperative decision making in social dilemmas. We study the effect of some frequently used strategies or guidelines (heuristics) on making decisions in situations of uncertainty and conflict. There is uncertainty because we do not know how each of the others will behave, and there is conflict because our own interest is at odds with the collective interest.

Evolutionary game theory has been used to study the viability of cooperation in a predatory world. While previous studies have helped to identify robust strategies, little is known about how success translates into the reproduction of cultural rules. Analogs of genetic replication may be deceptive if social learning and natural selection engender different population dynamics. I distinguished selection and learning based on whether rules are hardwired or softwired in the organisms that carry them. I then used genetic algorithms and artificial neural networks to operationalize the distinction. Applied for the first time to iterated Prisoner’s Dilemma, neural network experiments showed that researchers may need to be much more cautious in using Darwinian analogs as templates for modeling the evolution of cultural rules.

In this paper, rational social action is grounded upon a cognitive agent architecture. The notion of cognitive agent here referred to is defined as: (a) an ideal-type construct--not to be confused with the subjective, idiosyncratic individual as provided by the differential psychology; and (b) an AI-based notion, where the attention is drawn on the whole process that leads a system to acting, and on its internal makeup, i.e., the internal regulatory mechanisms and representations allowing a system to act adaptively in its environment. Inter alia, the following features of a cognitive agent architecture will be examined:

A computational instrument (DEPNET), calculating the network of dependence relations among agents in a common environment, and the formal theory of dependence upon which DEPNET is based, will be described. DEPNET will be shown to imply a very simple agent architecture, consisting of agents’ goals, actions, and resources. However, such a basic architecture allows a complex structure of agents’ interdependencies to emerge, and as a consequence, a variety of rational social (inter)actions to be predicted (influencing, exchange, and cooperation).

We examined dyadic interaction among eight archetypal social value orientations, or SVOs. (McClintock, 1978). These 8 SVOs differ in terms of two value systems: (1) concern with the well being of others and (2) concern with the well being of self (+,0 for Altruism; +,+ for Cooperation; 0,+ for Individualism; -,+ for Competition; -,0 for Aggression; -,- for Sadomasochism; 0,- for Masochism; and +,- for Martyrdom). All different pairs of SVOs (n=36) were formed, and each dyad played 60,000 2x2 games. For each game, the 2 payoffs in each cell of the matrix were generated randomly so that games had no particular structure (i.e., Prisoner’s Dilemma, Chicken, etc), or symmetry. The decision process was based 0on Kelley and Thibaut’s (1978) notion of transforming the given matrix into the effective matrix. That is, each SVO transformed the pair of given payoffs in each cell to a single number, or effective payoff (utility) based on its SVO (e.g., Cooperators took the sum of the payoffs to self and other; Altruists only considered the payoffs to the partner; Competitors subtracted the other’s points from their own). Next, each SVO chose the row or column with the largest average utility. We examined the long term outcomes of each SVO in two ways: (1) accumulated given points referred to as vitality, and (2) accumulated effective points (utilities) referred to as satisfaction. Vitality was completely determined by two things: (1) player’s concern with its own (given) outcomes, and (2) partner’s concern with player’s (given) outcomes. Although Individualism was the most vital SVO at the single person level, at the dyadic level, pairs of Cooperators were more vital than pairs of Individualists. For Satisfaction, no SVO did best overall. Satisfaction depended completely on the particular SVO pairing. The strongest mutual satisfaction was achieved by SVO pairs with identical effective matrices (e.g., Cooperator/Cooperator, Individualist/Altruist). Emergent Games: We also examined the long-term payoff matrices (the average of given points over 60,000 trials) produced by the pairings of different SVOs. A variety of such Emergent Games occurred, and the structure of each was tied perfectly to the SVO pair that formed it. Thus, the origins of structures are rooted in the social value systems of the involved parties. Notably, the Emergent Game produced by Cooperation versus Individualism is a classic Prisoner’s Dilemma.

The purpose of this article is to demonstrate that altruism, conceived as “generalized exchange, ”can evolve within a society of egoists. I define the “evolutionary giving game, ”in which each player holds some resource as its budget and can give any portion of it to any player. This game enables us to analyze the emergence of various kinds of social exchange, including generalized exchange. The results of computer simulations based on this game show the following. Though an unconditional altruist strategy is quickly beaten by nonaltruistic one, a kind of conditional altruist strategy, “in-group altruist strategy, ”will robustly evolve. The in-group altruist strategy dictates that the player gives only to “in-group members. ”Here in-group members are those who contribute sufficiently. Those who contribute sufficiently, but give even to “out-group members ”are also defined as out-group members. This strategy is consistent with in-group favoritism, observable in a variety of settings. The simulation results also show that the evolution of altruism based on this strategy is dependent on prospects of future contact, visibility, and risks.

Understanding social processes in which numerous agents are involved in iterated interactions is a difficult task. In the following I want to demonstrate that models based on rigorous simplifications are a promising approach for understanding complex social dynamics. The simplifications will go as far as to assume that individuals are living on a line interacting with neighbors to the left and to the right (Lineland). Another rigorous simplification will be that individuals are living on a checkerboard, interacting with their neighbors in the north, south, east and west (Flatland).

Research on social dilemmas and related problems are currently witnessing new, encompassing developments in theory and methods. In this paper, we will focus on one specific aspect of those developments—i.e., development of the “selective play ”paradigm. The attention of social dilemma researchers has traditionally been focused on isolated dyads or groups, and the wider context in which dyads or groups are embedded has traditionally been ignored. In this traditional research paradigm, players are implicitly assumed to be locked in a particular dyad or a group, having no chance of moving out and joining another dyad or group. This traditional research paradigm may be called the “forced play ”paradigm (Hayashi, 1995a) in the sense that players are “forced ”to interact with particular partners. Some of the interdependent relations we observe in the real world may be of this kind. A good example is the US-USSR relation before the collapsing of the Soviet Union. The two partners (the US and the USSR) are “forced ”to interact with each other no matter how strongly either party wanted to avoid interaction. However, many of the interdependent relations in the real world are not of this kind. Actually, only very few types of relationship (such as one between a parent and a child or between two siblings) are forced ones in this sense. Most personal as well as formal relationships we deal with in everyday life come with the possibility of exit. In our life, we typically choose to interact with our partners, and each has at least a partial freedom to leave the relation. Even the most long-lasting relationships such as marriage and friendship in the most contemporary societies entail possibilities of termination by voluntary moves of the people involved. People often form relations and leave them looking for better alternatives. Although the exit option was included in some of earlier studies of social dilemmas (e.g., Marwell & Schumitt, 1975; Orbell, Schwartz-Shea & Simmons, 1984) it was in the late 80’s that social dilemma researchers, though small in number yet, started systematic research efforts to explore theoretical implications of the option for leaving the current relationship and choosing a new partner. Resulting research paradigm may be called the “selective play ”paradigm (Orbell & Dawes,1991; Hayashi, 1995a). The purpose of this chapter is to explore theoretical implications of this emerging new paradigm. An additional purpose of this chapter is to provide a chance for the English speaking reader to be exposed to a literature on selective play published or reported in Japanese.

We introduce principles of self-organization in a dynamic individual oriented model of the evolution of friendship networks. The main aim of the model is to explain the emergence of structure in a friendship network from an initial situation of mutual strangers.

In contrast to the individual behavioral rules in previously developed models, we explicitly deal with the interplay between network dynamics and changing characteristics of individuals as a result of their friendships. Moreover, learning mechanisms and ‘bounded’ rationality principles, due to limited information and restricted capabilities to foresee strategically, are more explicitly dealt with.

The incorporation of these elements is facilitated by the application of principles of self-organization known from the fields of cellular automata and neural networks.

The present study simulates an enduring intergroup conflict in which each one of two interacting groups is represented by a unitary representative elected for a given constituency period. We assume that the conflict between the two groups can be modeled as an iterated Prisoner’s Dilemma game played by the groups’ representatives. In addition, we assume that the performance of each representative influences her constituents and that this, in turn, affects her prospects to be reelected. At the end of a constituency period, new elections are called for, and their results determine whether she remains in her position or is replaced by another representative. Our main objectives are: (1) to investigate the effect of this common democratic procedure, namely, the periodic election of group representatives, on the evolution of cooperation between the groups; and (2) to investigate the effect of the frequency of elections in the two groups on the evolving intergroup relations. Results of 150 simulations yield the following main results: (1) the dynamics of the intergroup conflict evolve into phases of welldefined patterns. (2) when mutual cooperation emerged, it was more enduring for high than for low rates of cooperation between the groups’ representatives; and (3) the prospects for reaching phases of mutual cooperation along the evolutionary path were considerably better for longer election periods, than for shorter ones.