The problem of social coordination is widespread across species and contexts, ranging from group cohesion and movement to courtship, aggression, and cooperation (see chapters on “Pro-social Behavior,” “Group-Spacing and Coordination,” and cooperation for more information on these topics). The primary similarity across these topics, uniting them under the umbrella term “coordination,” is that the outcome is due to joint action by two or more individuals at the same time that results in a desired outcome. Coordination is often used more or less interchangeably with cooperation in biology; however in economics, coordination specifically focuses on how individuals synchronize their behavior to reach (or not) a common goal (Guastello and Guastello 1998). It is often studied using coordination games, in which subjects benefit by coordinating their responses on the same outcome. These games are part of a set of tasks used in experimental economics to explore decision-making outcomes in humans. Aside from being useful to better understand human decision-making, the games are particularly amendable to comparative research (Smith et al. 2018) and thus can provide important insight into the evolution of decision-making, including coordination (Brosnan et al. 2013). We focus on two games, a coordination game called the Assurance game (or Stag Hunt game) and the closely related Hawk-Dove game (or Chicken game), an anti-coordination game.
Economic games take complex decision situations, such as coordination, and break them down to very simple choices. These choices are often dichotomous, although they can be more complex. The experimenter can also manipulate the paradigm to explore how decisions change in different contexts or within different social environments or alter the payoffs to explore how different types of decisions compare (i.e., coordination vs. anti-coordination). Due to their simplicity, these games can be easily adapted for multiple species and for use with different types of tasks, for example, either a manual task (e.g. choosing which of multiple tokens to return) or a touch screen or joystick-based computer task (e.g. choosing by selecting one of multiple icons on a screen). This is a key advantage for comparative work, since even basic morphological differences between species can strongly impact the degree to which the same task can be used with them, making these games particularly well suited for studying the evolutionary trajectory of decision-making. Although the games themselves are artificial, they are an excellent model system that can be used to derive hypotheses that can then be tested in more species-specific paradigms (Smith et al. 2018).
A key difference, of course, is that we cannot explain to nonhumans how the games work. This is typically solved in one of two ways; either they are allowed to learn the payoffs as they learn the task, or they are trained on the payoffs and then researchers explore whether their decisions change across different contexts. The former can be particularly useful in learning how subjects figure out what “game” they are playing, what the payoffs are, and how their decisions change as they learn the parameters. This is also relevant for understanding how people interact in non-laboratory situations. Although in the laboratory we can tell people the exact parameters of a decision task, in real life, individuals must figure out both the structure of the interaction and what the payoffs will be, the latter of which are often not foreseeable at the outset. Of course, this also means that we cannot be certain that subjects understand the task as we intended. As a result, it is important to “back-test” the adapted games that we use with animals on humans, using the same procedures that the other species received. Aside from ensuring that the humans respond to the modified game as expected, which indicates that the changes did not impact the goal of the procedure, it also verifies that any differences between humans and other species are due to differences in how they perceive or engage with the game, rather than differences in the procedure, and that similarities are really due to similarities in response, and not a result of having inadvertently made the game easier for one of the species.
Assurance Game (Stag-Hunt)
The Assurance game (also referred to as the Stag-Hunt game) is an example of a simple coordination game with two pure strategy Nash equilibria (NE; strategies whereby no player can do better by changing their strategy if their partner’s play remains consistent). The game was originally presented as a story by Jean-Jacques Rousseau and later modeled to a social dilemma (Skyrms 2003). In the story, two hunters can work together to take down a stag (a large prize), or they can individually hunt a hare (a much smaller prize). The dilemma is that while both individuals benefit more from working together to get the stag, if their partner does not participate, those that try for the stag outcome end up with nothing. However, they are always able to hunt the hare on their own. Thus, the dilemma is that participants both benefit the most from working together, but if they do not trust their partner, or think that their partner may not understand the task, they should make the lower-paying, but safe, choice to hunt hare.
This game is little studied in humans, likely because if the game is explained as we have just done, or a payoff matrix is presented, it is obvious that there is a mutually beneficial, high-paying outcome, which is what people overwhelmingly play. Therefore, most research studying cooperative decision-making in humans has focused on the Prisoner’s Dilemma, a cooperation game in which there is a high temptation to defect (for a better payoff), unless individuals play with the same partner repeatedly, in which case the gains from long-term cooperation outweigh the (one-time) gain from defecting on a cooperating partner, that will likely be followed by mutual defection. However, even in an Assurance game, if you do not know the payoffs ahead of time, and must learn them as you play the game, coordination may be harder to find than the human results imply. As mentioned above, this is true not only in the tasks used with nonhuman species, but often in real life. Indeed, Brian Skyrms, in his influential book on the Assurance game, has argued that it is a better model of many instances of human decision-making and deserves more attention (Skyrms 2003).
Researchers have ultimately tested five species representing a wide phylogenetic range within the primate order. This included humans, as well as three other primates that are highly cooperative and social and do well on a variety of cognitive tasks; chimpanzees (Pan troglodytes), rhesus macaques (Macaca mulatta), and capuchin monkeys (Sapajus [Cebus] apella). Chimpanzees are often used as a comparison with humans as they, with their sister species bonobos, are the extant great apes most closely related to us. Bolivian squirrel monkeys (Saimiri boliviensis) were also tested, which are closely related to and live in similar habitats as capuchin monkeys (both New World primate species) but do not show extensive cooperation. This allowed a determination of whether cooperation in daily life enhanced the ability to coordinate in this task.
The overall results showed clearly that some pairs of all of these species, including squirrel monkeys, coordinate on the payoff-dominant NE (Stag-Stag) in some contexts. However, there was substantial variation in how they did so (Brosnan et al. 2011, 2012; Vale et al. 2019). The New World monkeys, the most distantly related to humans, did the least well on the task. Three out of the five pairs of squirrel monkeys tested settled on the payoff-dominant NE, but while these pairs’ results were statistically significant, anecdotally, their frequency of choosing the Stag-Stag outcome was lower than that of the other species who found the NE, possibly due to the fact that squirrel monkeys do not appear to cooperate in the wild to the same extent as the other species tested (Vale et al. 2019). Capuchin monkeys also showed low levels of success in a manual task, but all subjects found the payoff-dominant outcome in the sequential computer task. Rhesus monkey and chimpanzee pairs were able to coordinate on the payoff-dominant NE in both the simultaneous and sequential versions, as were humans, although chimpanzees were highly variable.
These studies also provided insight into the mechanisms subjects used to solve the task. Capuchin monkeys were only able to solve the task when they could see their partner’s choices and were presumably using a matching strategy, with a bias towards the option that payed them better. Capuchins live in small, fairly cohesive social groups, and thus coordination likely occurs primarily when subjects are in close enough proximity to one another to see each other’s choices (Fragaszy et al. 2004). Capuchins therefore may have never experienced selective pressure for a mechanism for coordination other than behavioral matching or monitoring. Rhesus monkeys and chimpanzees (as well as humans) live in much larger social groups that frequently find individuals at much greater distances, or out of view. For example, chimpanzees coordinate behavior during group hunts and territorial patrols when they are out of view of one another (Boesch and Boesch 1989; Mitani and Watts 2005). This presumably provided selective pressure for the ability to remember partners’ past patterns of behavior even when they are out of sight.
Additional studies demonstrated that even the consistently successful rhesus monkeys were using a mechanism that was different than the humans’. In a control task in which subjects played a computerized simulation that was programmed to play different rates of Stag, humans probability matched, playing Stag at roughly the same levels as the simulation. Rhesus, on the other hand, played Stag almost no matter what the simulation played, indicating that they had simply formed a preference for this (typically) high-paying option (Parrish et al. 2014). This is particularly interesting as rhesus monkeys are known to probability match in other contexts, reinforcing that just because a species can use a particular cognitive mechanism does not mean that they will do so in every case; mechanism cannot be assumed from outcome. Finally, chimpanzees showed a high degree of variability, with chimpanzees that had extensive experience with cognitive tasks showing evidence of maximizing strategies that they could apply to novel situations. Chimpanzees who had very little experience with cognitive tasks, however, tended to match their partner, but with no bias towards mutual Stag play, or show no strategy at all (Hall et al. 2019). Although matching was still a relatively high-paying option, they would have done much better by coordinating on Stag. These chimpanzee results are hard to explain, given their behavior in the wild, and suggest the possibility that they were either failing to understand the task or were not motivated by it.
Humans showed one other curious outcome; in the manual task, many of the pairs settled on the risk-dominant Hare-Hare outcome. Looking at the data more deeply, however, it turns out that these pairs did not explore the parameter space and had never experienced the payoff-dominant Stag-Stag. It appears that after finding the Hare-Hare outcome, these subjects thought that they had solved the task and did not explore further. Language may be particularly relevant for humans in this context; in neither task were subjects explicitly told not to speak, but only in the computerized task did all pairs do so. Intriguingly, pairs that spoke about the game found the coordinated Stag NE, whereas pairs that never talked about the game settled on the mutual Hare outcome and, again, did not explore beyond this. These results may indicate that humans are using language to facilitate exploration of the parameter space.
As mentioned above, while the economic games are valuable for cross species comparisons, additional information can be gleaned from using tasks that take into account more species typical behaviors. In another series of studies, chimpanzees and children were presented with a paradigm that closely mimics the original Stag-Hunt story on which the Assurance game is based (Bullinger et al. 2011; Duguid et al. 2014). Pairs could individually forage on a low-value food (Hare) or move to a high-value food (Stag) that was only obtainable if they coordinated with their partner. Once they left the Hare option though, they were not able to return to it if their partner did not show up for coordination on the Stag option and were thus unable to acquire any additional rewards.
As in the economic games, both chimpanzees and human children were able to successfully coordinate on the mutually beneficial Stag-Stag option at very high rates. Both species primarily relied on a “leader follower” dynamic, where one subject would make their decision based on what the other did. Interestingly, when the experimental setup was changed so that subjects were not as easily able to see their partner’s decision (similar to the simultaneous version described earlier), rates of coordination decreased for the chimpanzees but remained high for the children. The researchers evaluated rates of communication between the species and found that children were able to maintain higher rates of successful coordination than chimpanzees in this more difficult scenario by increasing their verbal communication to replace their prior visual monitoring. Together with the humans’ outcomes in the Assurance Game prior game task, these results emphasize the importance of communication for humans in coordinating outcomes.
Hawk-Dove (Chicken) Game and Snowdrift Game
Sometimes individuals may be forced to confront a scenario where anti-coordination is required. This dynamic is captured best by the “chicken” game (made famous in the final car scene in the movie Grease). If two individuals are approaching one another head on, both individuals do best if one of them yields (to avoid injury), but the optimal choice for each is to maintain their course while the other yields (so they gain status). The optimal decision thus depends on doing the opposite of what the other individual is doing. This game has been studied extensively in behavioral ecology as the Hawk-Dove game, which models resource conflict but reflects the same payoffs as the chicken game. Whereas in a coordination game the Nash equilibria are for both players to coordinate on the same strategy, in an anti-coordination game the NE is for players to play different strategies. The Hawk-Dove game has three NE, two pure strategy, asymmetric NE where each player plays one of the options (the other plays the opposite strategy; bottom left and top right quadrants in Fig. 1b), and a mixed equilibrium where the players switch between these two pure strategies. The mixed equilibrium is the best equally paying solution because it maximizes both players’ benefits.
In a manual version of the Hawk-Dove game, using the token-exchange paradigm described above for the Assurance game, humans, but not capuchin monkeys or squirrel monkeys, were able to settle upon stable NE solutions (Brosnan et al. 2017; Vale et al. 2019). However, in a computerized version of the Hawk-Dove game (again using the same experimental methodology as described for the Assurance game), capuchin monkeys and rhesus monkeys settled on an asymmetric NE, although they were only consistently able to do so during the sequential version, in which they could see their partner’s choices (Brosnan et al. 2017). The difference in capuchins’ play between the two methodologies is likely due to the fact that aspects of the computerized task may enhance learning; because computers are faster than humans, the reward more rapidly follows the choice, there are more trials per session, and despite all of the controls in place during the manual tasks, there are presumably fewer distractions in computerized testing. Humans performed notably better than the other primates, with nearly half the pairs finding the mixed equilibrium, in which the players alternated between the two asymmetric NE. Indeed, this result is notable because unlike the other primates, who maintain the same level of finding the NE or do so in fewer circumstances in the Hawk-Dove game, the humans do much better in this more difficult game. One hypothesis is that, unlike in the Assurance game, where the coordinated Hare outcome may have seemed like a good solution in the absence of full information, the variability in this task may have meant that humans did not think that they had found a solution and were therefore more likely to explore.
A closely related game is the Snowdrift game (see chapter Snowdrift Game), modeled after a scenario in which two cars are prevented from getting home by a large snowdrift. The drivers both want the snow to be shoveled and could work together to do so, but both drivers would prefer that the other driver does it on their own. If the other driver does not do the work, however, the first driver should shovel the snow because the cost is less than the benefit of getting home (but does allow the partner to free ride). To test this, chimpanzees were presented with a situation in which they could work together to pull in food rewards or wait for their partner to pull them in (they received the same amount of food whether or not they helped). If neither individual pulled, the rewards were lost to both. Chimpanzees generally coordinated, pulling together with the partner, although they did seem to strategically minimize their effort by waiting longer to pull when the effort required for pulling was greater (Sánchez-Amaro et al. 2016).
Economic games are extremely useful for both determining how outcomes are similar, or not, across species and the underlying mechanisms behind them. Future research should focus on expanding the species tested, in particular testing species that are not known to coordinate extensively in natural contexts, to determine the degree to which individuals can find coordinated outcomes even if they have not been in environments that selected for it, or whether such an evolutionary history is critical for developing coordination. In addition, as we discussed above, the economic games are an excellent model that allows for direct comparisons across species, but they lack ecological validity. It will be important to continue to test the hypotheses that emerge from these games using more species-specific designs that allow us to explore how ecological pressures may have influenced species differently.
There are also other factors that may be influencing behavior in these games, such as hormones, social partner identity, or previous experience with cognitive testing. This will be particularly important to understand how these factors influence behavior across species. For instance, while previous research has demonstrated impacts of oxytocin on behavior in economic games in humans (Dreu et al. 2010; Rilling et al. 2012), recent work in capuchin monkeys finds no such effect (Smith et al. 2019). Social partner may also play a key role, as individuals should coordinate differently depending on their relationships, and as our chimpanzee results demonstrate, previous experience with cognitive and behavioral testing may be critical. Finally, we need data from a wider range of coordination games. The Battle of the Sexes game would be particularly relevant, as it is a coordination game with conflict (as opposed to an anti-coordination game like the Hawk-Dove game), where individuals do best to coordinate together on one outcome or the other, but there is disparity between the player’s individual preferences for each outcome. The addition of a wider range of species and modeled scenarios such as this would improve our understanding of the evolution of coordinated behavior and social decision-making in general.
- Brosnan, S. F., Parrish, A., Beran, M. J., Flemming, T., Heimbauer, L., Talbot, C. F., … Wilson, B. J. (2011). Responses to the Assurance game in monkeys, apes, and humans using equivalent procedures. Proceedings of the National Academy of Sciences, 108(8), 3442–3447. https://doi.org/10.1073/pnas.1016269108.CrossRefGoogle Scholar
- Brosnan, S. F., Wilson, B. J., & Beran, M. J. (2012). Old World monkeys are more similar to humans than New World monkeys when playing a coordination game. Proceedings of the Royal Society of London B: Biological Sciences, 279(1733), 1522–1530. https://doi.org/10.1098/rspb.2011.1781.CrossRefGoogle Scholar
- Brosnan, S. F., Beran, M. J., Parrish, A. E., Price, S. A., & Wilson, B. J. (2013). Comparative approaches to studying strategy: Towards an evolutionary account of primate decision making. Evolutionary Psychology: An International Journal of Evolutionary Approaches to Psychology and Behavior, 11(3), 606–627.CrossRefGoogle Scholar
- De Dreu, C. K. W., Greer, L. L., Handgraaf, M. J. J., Shalvi, S., Kleef, G. A. V., Baas, M., … Feith, S. W. W. (2010). The neuropeptide oxytocin regulates parochial altruism in intergroup conflict among humans. Science, 328(5984), 1408–1411. https://doi.org/10.1126/science.1189047.CrossRefGoogle Scholar
- Duguid, S., Wyman, E., Bullinger, A. F., Herfurth-Majstorovic, K., & Tomasello, M. (2014). Coordination strategies of chimpanzees and human children in a Stag Hunt game. Proceedings of the Royal Society of London B: Biological Sciences, 281(1796), 20141973. https://doi.org/10.1098/rspb.2014.1973.CrossRefGoogle Scholar
- Fragaszy, D. M., Visalberghi, E., & Fedigan, L. M. (2004). The complete capuchin: The biology of the genus Cebus (1st ed.). Cambridge, UK/New York: Cambridge University Press.Google Scholar
- Hall, K., Smith, M., Russell, J. L., Lambeth, S. P., Schapiro, S. J., & Brosnan, S. F. (2019). Chimpanzees rarely settle on consistent patterns of play in the Hawk Dove, Assurance, and Prisoner’s Dilemma games, in a token exchange task. Animal Behavior and Cognition, 6(1), 48–70.CrossRefGoogle Scholar
- Parrish, A. E., Brosnan, S. F., Wilson, B. J., & Beran, M. J. (2014). Differential responding by rhesus monkeys (Macaca mulatta) and humans (homo sapiens) to variable outcomes in the assurance game. Animal Behavior and Cognition, 1(3), 215–229. https://doi.org/10.12966/abc.08.01.2014.CrossRefGoogle Scholar
- Rilling, J. K., DeMarco, A. C., Hackett, P. D., Thompson, R., Ditzen, B., Patel, R., & Pagnoni, G. (2012). Effects of intranasal oxytocin and vasopressin on cooperative behavior and associated brain activity in men. Psychoneuroendocrinology, 37(4), 447–461. https://doi.org/10.1016/j.psyneuen.2011.07.013.CrossRefPubMedGoogle Scholar
- Smith, M. F., Watzek, J., & Brosnan, S. F. (2018). The importance of a truly comparative methodology for comparative psychology. International Journal of Comparative Psychology, 31(0). Retrieved from https://escholarship.org/uc/item/6x91j98x.
- Smith, M. F., Leverett, K. L., Wilson, B. J., & Brosnan, S. F. (2019). Capuchin monkeys (Sapajus [Cebus] apella) play Nash equilibria in dynamic games, but their decisions are likely not influenced by oxytocin. American Journal of Primatology, 81(4), e22973. https://doi.org/10.1002/ajp.22973CrossRefGoogle Scholar