The Thoughtful Animal

i-2c8a01fe9d63140bf53a83d1a85d49c3-dog cooperation cartoon.png

ResearchBlogging.orgCooperation and conflict are both a part of human society. While a good deal of the academic literature addresses the evolutionary origins of conflict, in recent years there has been an increased focus on the investigation of the evolutionary origins of cooperative behavior. One component of cooperative behavior that might be present in other animals is aversion to inequity. Some scientists have suggested that inequity aversion may itself be the main factor driving the enforcement of cooperation. Put simply, inequity aversion is the resistance among partners to unequal rewards following identical effort, and therefore requires an individual to be aware of his or her own efforts and payoffs as well as those of his or her partners. For example, non-human primates (such as capuchin monkeys, chimpanzees, and cottontop tamarins) will refuse to participate in a cooperative problem-solving task if their partners receive a better reward for the same effort.

But what about other cooperative animals? Canids are known to hunt cooperatively, for example. And while the process of domestication has facilitated human-dog social interaction at the potential expense of dog-dog social interaction, dogs clearly engage in cooperative behaviors, generally. While dogs readily cooperate with humans (e.g. guide dogs, drug-sniffing dogs), it is often the case that multiple dogs must cooperate with each other in order to work with a human toward a common goal (such as with hunting dogs). Since dogs show understanding of human intentional communicative cues, and since dogs are sensitive to human behaviors directed towards them as well as towards others, they are a perfectly suited species with which to investigate the question of inequity aversion.


Are pairs of domesticated dogs sensitive to the inequity of rewards received from a human experimenter for identical actions? In this study, the experimenter requested a specific behavior from each of two dogs, one after the other, in order to receive a food reward. In all cases, the dogs were familiar to each other, and all dogs had been trained to give their paw to their owners upon request. In each trial, the experimenter avoided any communication with either dog, beyond the verbal request give their paws. A total of twenty-nine dogs participated.

dog fairness 4.jpg

Figure 1: The experimental set-up. The owner stood behind the dogs, while the experimenter asked the dogs for their paws, while avoiding eye contact.

The conditions were as follows:

  • ET: “Equity test,” where both the subject and the partner received the same low-value reward. This condition served as baseline.
  • QI: “Quality inequity test,” where the subject received a lower-value reward than the partner.
  • RI: “Reward inequity test,” where the subject received no reward, while the partner received the low-value reward.
  • EC: “Effort control,” where both dogs received the low-value reward, but the partner did not have to perform the requested behavior in order to receive that reward. This condition was included to ascertain whether the results from the inequity conditions could be the result of the subjects’ awareness of the effort expended by the partner.
  • “Assessment,” conducted with each dog without a partner present, to assess whether the dogs would give their paws to an unfamiliar experimenter thirty times.
  • NR: “No-reward,” was conducted following the assessment condition, without a partner present, to see how long the dogs would continue giving their paws without a reward.

In all conditions, including the non-social assessment and no-reward conditions, both the high-value rewards (sausage) and the low-value rewards (bread) were present and visible to each dog. Each testing session included up to thirty trials, in which the partner performed immediately before the subject. If the subject refused to participate prior to the completion of all thirty trials, then there would be fewer trials conducted. The experimenter would repeat the command up to ten times before aborting the rest of the session. Further, all conditions were randomized and counterbalanced for each subject, with the exception that the first condition was never the “no-reward condition” or the “reward inequity” condition. The non-social assessment control was conducted before the social conditions for half the subjects, and after the social conditions for the other half of the subjects.

i-22650821465b4181d9ac2d64a168778b-fairness dogs1.jpg

Figure 2: Results for the four social conditions as well as the non-social no-reward condition, showing the number of trials the paw was given on command. Long bars are median values, boxes show middle 50%.

As would be expected if the dogs had inequity aversion, the subjects refused to give their paws to the experimenter sooner in the “reward inequity” (RI) condition compared with the baseline equity (ET) condition. Further, the presence of the partner was critical: subjects stopped responding significantly earlier if a reward-receiving partner was present than in the non-social “no-reward” condition.

Recall that the experimenter would repeat the verbal command up to ten times before ending the testing session. If it took more requests before they would respond, it could reflect the subjects’ hesitation.

i-e4ca0dc6793bdbf07fab8506b0cfadca-dogs fairness2.jpg

Figure 3: Results for the four social conditions as well as the non-social no-reward condition, showing the number of verbal prompts before the behavior was executed. Long bars are median values, boxes show middle 50%.

The only significant difference was that in the reward inequity (RI) condition, subjects were less willing to participate (it took more verbal commands), when compared with the baseline (ET) condition. The dogs were more also hesitant to give their paws when their partners were rewarded for the same action (RI) than during the non-social no-reward (NR) condition. This again suggests that the presence of the partner is critical, since in both of these conditions, the subject did not receive the reward.

In addition, the experimenters recorded a number of other variables related to the subjects’ stress during the testing sessions, defined as the number of scratches, yawns, licks, and avoidance of the gaze of the partner, per trial. This comparison indicated that subjects were more distressed in the reward inequity (RI) condition compared with the baseline (ET) condition and the non-social no-reward (NR) condition. As with the previous results, this finding suggests that the presence of the partner is key: not getting a reward in the presence of a rewarded partner is more stressful than not getting a reward while alone.

Taken together, these results indicate that dogs are sensitive to the unequal reward distribution, rather than simply a difference in cost (effort) or benefit (reward quality).

However, it is possible that the dogs were reacting simply to the presence of the other dog, rather than the reward distribution. To account for this, a second experiment was conducted to control for the presence of the partner. The set-up was the same, and the conditions (randomized and counterbalanced for each subject) were:

  • SC: “Social control,” where both the subject and the partner received no reward.
  • RI: “Reward inequity test,” where the subject received no reward, while the partner received the low-value reward. (the same as the RI condition in the first experiment)

dog fairness 3.jpg

Figure 4: Results for experiment 2. Number of trials the paw was given (left), and number of verbal commands required before the behavior was executed (right). Long bars are median values, boxes show middle 50%.

Overall, the dogs hesitated longer (more verbal commands were required) and also displayed more behaviors indicative of distress in the RI condition compared with the SC condition. These results support the earlier conclusion, that the subjects were reacting to the unequal reward distribution, rather than simply to the presence or absence of another dog.

Taken together, these results suggest that dogs possess a sense of fairness; they refuse to obey a verbal command if they are unrewarded in the presence of another dog receiving a reward for the same action. But these results differ slightly from the inequity aversion studies of non-human primates. For example, non-human primates, in addition to being sensitive to unequal reward distributions, are also sensitive to the more subtle differences in the quality of the reward and the effort expended by their social partners. While all owners reported that the dogs preferred the high-quality reward (sausage) over the low-quality reward (bread), it is possible that the dogs were so used to be trained that they were eager to obey the commands, even in exchange for the lower-quality reward. It is also possible that the dogs’ general eagerness to please humans explains the results with respect to the quality of the food rewards. Given these potential confounds, we can only speculate as to the reason for their apparent indifference to the quality of the rewards.

The researchers suggest, perhaps, that the psychological mechanism required for an aversion to inequity requires the ability to perceive a relation between relations – that is, to compare the relation between one’s own effort and reward, with the relation between the partner’s effort and reward. Only humans, chimpanzees, and baboons are known to be able to solve tasks that require this level of relational complexity. It is therefore possible that dogs simply lack the ability to comprehend such “relations between relations.” This could explain the results from this study, wherein the dogs were sensitive more generally to being rewarded or not, but not to the more nuanced inequities. The authors wonder if the “sensitivity toward being rewarded or not [is] the precursor to cognitively higher-level forms of inequity aversion.”

Another significant difference between these results and those from studies of non-human primates is that the dogs never rejected food. The dogs only responded to the inequity when they were in the disadvantaged position (not being rewarded); it was never the case that the rewarded dog refused the reward. The dogs were not willing to pay a cost in rejecting unfair advantages, even when it was in their favor, while some non-human primate species are willing to pay such a cost. The more sophisticated form of inequity aversion in non-human primates (as well as humans) may indicate another fundamental difference between dogs and primates, perhaps suggesting again that dogs possess a precursor to more complex forms of inequity aversion; a sort of proto-fairness.

This raises several questions. First, is this more basic form of inequity aversion unique to primates and domesticated species, or can it be found in other species that are known to act cooperatively in their natural environments? Second, does the insensitivity to food quality and effort seen in dogs reflect their domestication, or developmental training, by humans? To address the first question, other social species – domesticated and non-domesticated – should be tested in similar experiments. To address the second question, wild and hand-reared wolves, as well as other domesticated canids (such as the domesticated foxes we’ve come to know and love) should be tested in this kind of task. If wolves’ performance on this task more closely mirrors the non-human primates and domesticated foxes behave similar to domesticated dogs, then perhaps there was a tradeoff during the process of domestication wherein some of the sensitivity to reward inequity among conspecifics was lost, as sensitivity to human social cues was sharpened.

Range F, Horn L, Viranyi Z, & Huber L (2009). The absence of reward induces inequity aversion in dogs. Proceedings of the National Academy of Sciences of the United States of America, 106 (1), 340-5. PMID: 19064923

Comments

  1. #1 Charles Sullivan
    September 20, 2010

    “it was never the case that the rewarded dog refused the reward.”

    Why is that not surprising?

  2. #2 micha
    September 21, 2010

    Was it a “partner inequity test”? Or was it a “this guy is arbitrary, why expect anything from him” test?

    -micha

  3. #3 muddy
    September 21, 2010

    All treats are high value treats, at least to my dogs.

  4. #4 Neuroskeptic
    September 22, 2010

    Hmm, it’s a nice study, but the results seem messy: the differences between the RI and the NR conditions, which is the crucial comparison to prove that the dogs care about fairness, seemed pretty small, with lots of overlap.

  5. #5 The Dog Zombie
    September 23, 2010

    The dogs’ reactions to quality inequalities is really interesting! Two things come to mind. a) Were the investigators sure that the sausage was actually a higher reward for all the dogs than bread? It seems obvious that it would be, but you never know. (Finding the right reward for a particular dog can be a problem for some trainers/owners.) Hopefully they tested that in some way. b) I wonder if never refusing food is a scavenger trait. I don’t think a scavenger species has been tested on this particular test before. Who knows.

    Thanks for reviewing this article — fun read.

  6. #6 Paul W., OM
    September 23, 2010

    I’ve got to wonder how much these results would vary given differently socialized dogs.

    It seems to me that dogs have two main opportunities for early learning about fairness and greed—when eating, and when playing.

    Some people let their dogs manifest dominance relations over food, some don’t.

    I feed my dogs from separate dishes, and don’t let them eat from each other’s dishes. (A lot of people do that.)

    A friend of mine used to just leave food out for a while, so that all the dogs could get their fill, then remove the food, so that they couldn’t eat all day. The dominant dog would eat all it wanted before letting the next dog have any, and so on in a strict pecking order. Even so, in practice all the dogs got to eat all they wanted, once per mealtime, so they may have developed expectations like “everybody gets to eat every time, and nobody goes hungry.”

    In a rather more natural scenario, where there often isn’t enough to go around, dogs might learn that you should get what you can when you can, and that you just shouldn’t expect such things to be anything like equitable.

    Even in that case, what a dog might learn is to fight against inequity—if you don’t fight for food, you’ll go hungry a lot.

    Even in that harsh scenario, there’s an opportunity for social learning about a weak sort of fairness—e.g., don’t let the dominant dog eat all the food, and don’t expect a submissive dog to be completely submissive. (If you let them have some food, they’re likely to let you eat more food without a fight.)

    Similar considerations apply to playing. In particular, dogs may learn from experience that if they don’t let the other dog choose the game or lead the pack or win the fight, sometimes, they end up with nobody to play with—being completely dominant is self-defeating. (To my chagrin, I seem to recall learning a similar lesson from experience as a child, maybe a couple of years later than I should have.)

    If there’s a sort of “universal grammar” of social relations, I’d be surprised if there weren’t a couple of crucial learned parameters relevant to these sorts of things, e.g., (1) whether to expect fairness at all, or just give up on the idea, and/or (2) how disproportionate rewards can be while counting as “fair enough” not to be offensive or not to be worth fighting over, and (3) what factors are expected to correlate with those proportions.

    (E.g., do relatively dominant dogs expect twice as much treats or play-fight winning as the next most dominant dog, and does that dog learn to accept that as inoffensive? Are there different parameters for food and for playing, or are they linked and tweaked?)

    It’d be interesting thing to see a thorough study of these things, with different populations of dogs raised differently. (E.g. fed fairly vs. unfairly vs. asocially, allowed to fight for dominance or prevented from doing so as much as is practical, etc.)

  7. #7 Paul W., OM
    September 23, 2010

    A couple of other things just came to mind.

    It seems basically unsurprising to me that dogs would have some notion of fairness, either innate or learned.

    After all, even small fish show prosocial behaviors and a reward/punishment scheme for prosocial/selfish behavior.

    (I’m reminded of “predator inspection” studies, where for some kind of fish or other it turns out that they more or less take turns checking out potential predators, and reward each other for taking that prosocial risk, and punish each other for failing to do so. It’s unclear to me whether this is a manifestation of a more general sense of “fairness” or “equity,” as opposed to some very specific instincts about predator inspection.)

    I have to wonder whether schooling fish have a strong sense of fairness too, and whether mammals’ sense of fairness is derived from that, with the basic mechanism being conserved for hundreds of millions of years.

    It may be that dominance-oriented animals evolved from non-dominance oriented animals, maybe schooling fish, and that fairness-consciousness is the basal state. Dogs being so authoritarian and greedy may be mostly a suppression of the ancestral condition, or a retuning of it by setting an ancient greed/fairness parameter toward the greedy end of the behavioral spectrum.

    Or something like that. Probably something more complicated. :-/ (Ob. caveat: I’m a layman about this stuff and speculating wildly.)

    I seem to recall that in the predator inspection studies, the fish had a strongly equitable, roughly tit-for-tat scheme. Every fish had to do the predator inspection scheme comparably often, or the others would shun it. (If it did its duty, it’d be favored by being allowed the favored position near the center of the school—the safest position. If it didn’t, it’d be excluded from the school and thus most vulnerable.)

    It seems reasonably likely to me that the basic social bookkeeping mechanism may have been conserved for hundreds of millions of years, and may be manifested in all sorts of ways in all kinds of vertebrates. (E.g. in seemingly asocial but territorial animals, who may remember intruders.)

    It also seems reasonably likely that a fairly general fairness schema may be highly conserved, but applied very differently and selectively in very different kinds of situations (e.g., respecting territory in asocial grazers vs. food sharing in social hunters).

    On the other hand, once you have the social bookkeeping mechanism conserved, it may be easy to re-evolve a lost facility for dominance hierarchies or equitable fairness with social enforcement. Computationally, the hard part would be keeping track of favors/offenses. Deciding what to do with that information—whether to exploit others, and when to allow yourself to get exploited—would be a comparatively simple high-level “policy” decision.

  8. #8 bo moore
    September 24, 2010

    How silly: these tests are based on human ideas of reward and inequality. To be accurate, this test would have to be designed by dogs. My dogs never complain about ANY food offered; they do ‘act out’ if one gets to go for a ride in the truck and one stays home. They respond to being left out of activities. They have different taste palates anyway: one prefers carrots and other vegetables, the other regular dog biscuits. These tests don’t account for taste!

The site is currently under maintenance and will be back shortly. New comments have been disabled during this time, please check back soon.