Insights from Game Theory for Canine Conflict

The Sequential Assessment Model of the Hawk/Dove game provides a mathematical basis for understanding canine conflict

After reading an amazing chapter in the book “Power in the Wild: The Subtle and Not-So-Subtle Ways Animals Strive for Control Over Others” By Lee Alan Dugatkin, I discovered the application of Game Theory to animal power dynamics. Power is universally a factor in everything animals do - from what they eat, where they eat, who they mate with, and how often, how many offspring they produce, what social groups they form & depose, the alliances they make, and so much more. Social power dynamics are observed in everything from insects to fish to birds to monkeys, mammals, and even us. Power can have biological effects on the body, social rank can affect things like what hormones are in the bloodstream, numbers of parasites one is subjected to, and in some species, power can even change their sex.

Hawk/dove comes from Game Theory, which is a mathematical framework for understanding, analyzing, and predicting interactions between rational “decision makers” given a strategic situation in which the outcome depends on the decisions of all the players. In Hawk/Dove, you have two players contesting a resource. The players may choose one of two strategies, they can be hawkish - and fight for the resource…or dovish, and retreat. If both players choose Hawk, a fight occurs, one player gets the resource and the other incurs the costs of losing. If one player chooses Hawk, the other Dove, the hawk wins the resource at no cost and the dove retreats. If both players choose dove, they split the resource. This is easily translatable to dogs, in the sense that a dog may engage in some ritualized aggression over a contested resource (such as growling or biting) and the other dog may choose to rise up to that conflict, or retreat. If neither dog is willing to engage in conflict, they may share.

The Sequential Assessment Model is an additional component that can be added to the Hawk/Dove game, which accounts for varying “levels” of conflict and de-escalation. It allows for an animal to assess and determine their rival’s intentions and willingness to escalate, allowing for conflicts to be resolved at the lowest possible cost. The outcome of these sequential hawk/dove games is dependent on both the value of the resource, and the cost of conflict. The higher the value, or benefit, of acquiring that resource the more beneficial it is to take a hawkish strategy. The higher the cost of engaging in conflict, the more beneficial it is to take a dovish strategy. An easy example is the scale ranging from nations posturing on social media, to all-out nuclear war, and everything inbetween. At a certain point, the costs of conflict outweigh the benefit, and such a strategy is not beneficial. Similarly, at the lower levels, the benefits of conflict outweigh the costs, so conflict ensues.

These insights from Game Theory allow us to visualize the progressive ritualistic behaviors a dog exhibits around resources (such as food, water, toys, space, social members, territory, etc) as a progressive assessment model involving hawkish or dovish strategies and their associated costs & benefits. For example, if we rank agonistic dog behavior on a scale of 1-10, 10 being an all out brawl, 1 being “whale eye” or an agonistic pucker of the lips, we can run a simulation between two dogs using the hawk/dove paradigm. A resource is contested, dog A and B choose a hawkish strategy at level 1 and both may freeze & stare. So, sequentially, the next option might be a growl. Both dogs choose hawk, so the assessment continues. Lets say at a level 3, Dog A air snaps, and Dog B determines that benefit of the resource is not worth the potential cost of further escalation, and Dog B doves. The level to which a rival was willing to escalate and the point at which they are no longer willing is valuable information in the quest for power. While not a perfect representation of the model, hopefully this gives you an idea.

Now, how can this help us with our dogs? Well lets say we have a dog that “does great at the dog park” and “is an angel at home”, but is “leash reactive”. A closer examination of this dog’s social interactions at the dog park or at home may indicate very “hawkish” behavior, contesting resources and escalating in the use of social pressure to maintain control of them. However, since the potential costs of escalating too much at point blank range with another dog may be too high, there is a limit to the level of escalation. Now, we take that dog on a walk. Lets say we walk this dog the same perimeter every day, multiple times a day, and the dog starts to perceive this area as their territory. On this walk, we also feed the dog treats to practice their loose leash walking. So this territory has proxy value, as a space in which food is accessible and comes from the human’s pocket consistently. So the benefit of these resources is high, and the dog has a history of taking a hawkish strategy. Now, suddenly, there is competition. A dog has invaded the territory (just by virtue of being on the same walking path). Now there is competition, the resource is contested. As a result of the leash, the dogs cannot engage in any close-range ritualistic behaviors, and must “project” their social pressure and willingness to escalate over long range. Additionally, because they are not close to each other, there is no “cost” to rapid escalation. These factors combined are a recipe for lunging and barking on sight. I believe this explains similar situations in which dogs are barking and growling at a gate, but when the gate is removed, they immediately stop. There is no cost for the hawkish strategy because of the barrier, and thus there is no reason to dove. Kind of like how people will say anything on Facebook, but won’t say it in person (haha).

Examining conflict from this angle also gives clues for how dominance hierarchies can be a “collective computation” of memories of certain individuals and how they behaved, remembering the extent to which some are willing to escalate, the points at which they de-escalate, what resources they consider to be high value, etc. constantly fluctuating as new experiences occur or social members are added or removed. A hawk/dove sequential assessment of dog reactivity and the inherent costs & benefits accrued implies that there may be something fundamentally wrong with how we are approaching these issues, which may account for the overwhelming lack of success stemming the tide of dog behavior issues despite the best insights from “learning theory”, and the slow rate at which these issues are resolved.

I hope you guys found this as interesting as I did. Please consider becoming a paid subscriber, it really helps motivate me to sit down and spend the time to type these thoughts out and listen to books on repeat trying to absorb all the information I can!