# Remembering the dangers of nationalism

Canadian veterans and the Royal Canadian Legion at Remembrance Day ceremonies on Lower Field of McGill University (November 11th, 2012). Photographed and arranged by Adam Scotti; reproduced with permission.

Remembrance Day is a time to reflect on past conflicts and honour the men and women that did not return home from war. The day commemorates the armistice signed on the morning on November 11th, 1918 to formally end the hostilities of the First World War “at the 11th hour of the 11th day of the 11th month”. It is observed by a minute of silence to honour the fallen in all armed conflicts. It is difficult for me to trace my genealogy past the Russian Civil War and draw any personal connection to the First World War. But like many Russians, war losses are a looming memory as none of my great-grandfathers returned home from the Second World War. However, even with careful memory of past conflicts, it seems that we are not capable of avoiding new ones.

In a typical high-school history text, one will see WW1 attributed to nationalism. This is a form of tag-based ethnocentrism, where the arbitrary tag is nationality — a slippery concept for formal definitions but one for which many of us have an intuitive grasp from cultural indoctrination. Of course, nationality seems like a very high-level tag, with many arbitrary distinctions possible to separate people within a single nation (to take a recently prominent one: political party allegiance). As such, our concerns about the expansion of the moral circle apply just as well to nationality as they do to all of humanity. If we understood how the in-group expands from a family, clan, or tribe to the nation, we would be most of the way to an emphatic civilization. The only hurdle would be to understand what the permanent absence of an out-group entails. For now, I will leave the issues of connecting tribe-level ethnocentrism to nationalism at the level of analogy.

A popular approach in understanding nationalism and war is to trace it to a potential evolutionary origin. Usually, this means understanding how tribal ethnocentrism and warfare could have emerged in late Pleistocene and early Holocene humans. Choi and Bowles (2007) take this approach through an agent based group-selection model. Agents form groups (or tribes) of three generations, with each generation consisting of 26 individuals. As far as I can tell, only one of the three generations reproduces and participates in interactions. Each individual is either an altruist or not (in-group strategy), and parochial or tolerant (out-group strategy). Within the tribe, a public-goods game is played with altruists cooperating and nonaltruists defecting. Reproduction is pseudosexual (individuals don’t have sexes, but are still paired to produce offspring) and proportional to the fitness of both parents with full recombination (the in-group and out-group strategies are at two independent loci) and a small probability of mutation and a small random migration rate between groups.

Between the groups, the authors employ a very complicated mechanism. All 20 groups are paired randomly to interact, each group has a probability equal to the group’s fraction of tolerant agents to choose a non-hostile interaction. If both groups choose to be non-hostile, then each tolerant agent gets a benefit $g$ from each tolerant agent of the other group. If either group is hostile, then there is a chance of war equal to the difference in proportion of parochial altruists (warriors) between the groups ($|\Delta_{ij}| = |f^{\text{PA}}_i - f^{\text{PA}}_j|$). If war does not occur, then there is still no benefit to tolerant agents, and the individual payoff is completely from the in-group game. If war occurs, then a constant fraction of parochial altruists perish (14%) regardless of if the war is a tie (with probability $\frac{1}{2}(1 - |\Delta_{ij}|$) or the side with more PAs is victorious. If the stronger group is victorious, then they kill a fraction of $2.5 |\Delta_{ij}|$ of the weaker group’s civilians. The authors do not make clear what happens when $2.5 \Delta_{ij} \geq 1 - f^{\text{PA}}_j$, but presumably there is a floor effect and every member of group $j$ is killed. The stronger group (both civilians and parochial altruists) then produces the offspring to repopulate the losing group.

In the above model there are two stable equilibria. In one there is about 15% of both in-group altruism, and out-group parochialism, and in the other there is 85% of each. In the first equilibrium there is very little war and hostility, in the second it abounds, but at levels that are not unreasonable given the archaeological data. Transitions can happen between these equilibria relatively quickly (around 200 cycles, or 5000 years for human generations). The long term average tends to populations that are either parochial altruists or tolerant non-altruists, with very little in-between. From this, Choi & Bowles (2007) conclude the co-evolution of parochial altruism and war.

This paper caused a stir when it came out, and has been heavily cited. From a modeling perspective, I think it suffers from numerous flaws (most introduced in the arbitrary and complicated war mechanism) and could be approached cleaner analytically, but I will save the details of my critique for a future posts. The main achievement of Choi and Bowles (2007) is an attempt to be more realistic that the abstract models typically studied in evolutionary game theory, and to reinforce the important point that hostility and altruism often go hand in hand. Especially in the case of ethnocentrism, it is important to remember the dangers in the cooperation it brings. As I wrote in 2010 in the context of a different model:

The evolution of ethnocentrism … is a double-edged sword: it can cause unexpected cooperative behavior, but also irrational hostility.

Choi JK, & Bowles S (2007). The coevolution of parochial altruism and war. Science (New York, N.Y.), 318 (5850), 636-40 PMID: 17962562