Saturday, July 4, 2009

Altruism through genocide

Did human morals and altruism arise from intergroup warfare?

One of the long-standing problems of evolutionary biology is how to account for altruism as a trait of human nature when the Darwinian mechanism is relentlessly individualistic. To sacrifice one's self for non-relatives is abhorrent to a dyed-in-the-wool Darwinist, and even towards relatives, the calculus yields rather modest sacrifices, rapidly attenuating with distance.

In "The Selfish Gene", Dawkins (following Williams) even proclaims the individual to be an epiphenomenon- the temporary abode of genes which are the true (and semi-immortal) objects of selection, to whom their habitation is a feeding and replicating robot. (One could make the same point about individual nucleotides within genes, to take this analysis to a reductio ad absurdum.) In contrast, Stephen Gould wrote a miasmic defense of multi-level selection that incorporated individuals, groups, and even species and higher groupings as objects of natural selection. But he was slightly ahead of his time, for group selection has only now begun to come back into the evolutionary literature in earnest.

It is to group selection we must turn to investigate the origin of moral traits, since morals (such as altruism) concern how we treat others, and especially how we moderate our selfishness in the service of larger groups. This is not a new story in evolution, since cooperation happens among bacteria, (mutualist associations among methanogens and methane oxidizers come to mind, as do other mixed species biofilms), plants (lichens are a simple example, as are corals), and insects (ants, termites, etc.), among many, many others. Grouping and cooperation has been a winning strategy many times over in the history of life, including the critical innovation of the eukaryotic consortium over its bacterial ancestors.

Darwin himself suggested this rationale in human evolution, proposing selection of groups with "a greater number of courageous, sympathetic, and faithful members, who were always ready to warn each other of danger, to aid, and defend each other ... would spread and be victorious over other tribes." (Descent of man, quote taken from the paper being reviewed). The problem has always been figuring out in quantitative population genetics terms whether this made any sense, especially how a strategy of altruism could survive against cheating from within the group as well as from gene dilution from outside sources. There are strictly individualistic rationales for some level of cooperation among organisms, but they never rise to the level of the ultimate sacrifice, as exhibited by cells in our body committing suicide rather than turn cancerous, or soldiers offering their lives on the battlefield.

Population genetics revolves around the concept of "fitness", which quantitates differential reproductive success. In common parlance, fitness refers to traits that enhance survival (thicker fur, higher intelligence, disease resistance, etc.), but as used in evolutionary biology, fitness refers to the end effect of trait differences, expressed in the proportion of the individual's genes in the next generation, divided by the proportion in the current generation. Neutrality would be a value of 1, and comparisons are usually expressed as differential fitness, between alleles of a gene, or between whole genomes. The author mentions that a fitness differential (or cost) of -0.03 would drive its occurrence from 90% of the population to 10% in 150 generations, illustrating how small differentials can have large effects over time.

The paper at issue, by Samuel Bowles ("Did warfare among ancestral hunter-gatherers affect the evolution of human social behaviors?"), is mathematical, creating a model of small populations with given amounts of gene flow from outside (or, conversely, inbreeding), given amounts of warfare in which groups extinguish other groups and expand accordingly, and given amounts of altruism, expressed as individual costs (death in warfare by warriors) and group benefits (success in warfare). Then the paper draws on various other sources to decide on reasonable bounds for these various values, plugs them in, and comes up with the finding that altruism in these simplified terms could be selected for, even at substantial individual cost.

The equation Bowles comes up with for his model is:
Where c is the individual fitness cost to altruists (set at 0.03, as above), k (kappa) is the probability per generation of being in a conflict with another group, L (lambda) is is the probability of group survival in such a contest, F(ST) is the inbreeding coefficient of the group (higher means more inbred), and n is the group size. The equality represents the boundary case, where increased c, the individual cost of altruism, breaks even versus its group benefits. These benefits are represented by L as arbitrarily adjusted (subscript A) by the presence of altruists, where an extreme value of 3.3 means that a group with 10% more altruists than its competitor suffers half the mortality in a contest. Bowles admits that there is no way to empirically estimate this parameter, and offers a curve of possible values, from 0 (no effect) to a high of 3.

Another key variable is delta, the overall rate of adult violent mortality (somewhat determinable empirically), taken to represent warfare in some proportion. k equals 2*delta, as a simplified estimate, assuming half the combatants win and half lose on average.

Bowles then estimates some semi-empirical values from the ethnographic and archaeological literature. First, for conflict-derived mortality (delta), he settles on 14% as an estimated rate of violent death in prehistoric cultures among all adults (note how vastly higher this is than what we experience in contemporary life). This is transformed to rates of conflict (k) and rates of defeat or victory in such contests (L), adjusted by the putative in-group altruism factor. Group sizes and inbreeding rates have been extensively studied elsewhere, so are not a special focus in this article. Group sizes are averaged at 26, (though an infinite group size is also treated), and several possible inbreeding coefficients (ratio of between-group genetic variance to total variance) are offered in the derived curves (below).

The graph shows solutions to this equation with these various values plugged in. The equation is satisfied and altruism favored for areas above and to the right of each curve.

"Wartime mortality (delta) and the effect of altruism on success in conflicts (L, or lambda A) sufficient for the proliferation of an altruistic trait with individual fitness cost (c)=0.03 for three estimates of the extent of genetic differentiation among groups (F(ST)). Shown are the values of c consistent with Eq. 6 for the estimated F values from (12, 17). The representative values of delta are from Table 2. Populations on the horizontal axis in italics are from the ethnographic sample; the rest are from the archaeological sample." (from Bowles, 2009)

So, granted that this is a quite artificial situation, and that I am no expert in the field, what this paper does is to establish inter-group selection as plausible and quantifiable mechanism in accounting for trait-based altruism in the form of warrior behavior and sacrifice. The empirical boundaries seem plausible, while some of the mathematical simplifications seem less so (in this model, each conflict results in extermination of one group and doubling of the other, for instance. On the other hand, all non-altruists get to survive if their group wins a war, while altruist warriors die with 20% probability). Specifically, the right-ward parts of the curves indicate that given realistic rates of death (delta) from conflict, altruism could be selected for despite substantial costs (c) and modest group benefits (L). There are many other possible rationales for selection of in-group altruism traits, so this setting of warfare should not be seen as exhaustive.

The analysis just codifies and works out in population genetic terms the basic insight that groups are quite evidently objects of selection. As long as there are genetically distinct human groups, and those groups are being killed by other groups, (richly detailed and cheered on in the old testament, in the epics of many other cultures, and recently witnessed in the Balkans, Rawanda, etc.), there is evolutionary selection of groups going on. One focus of this selection would then be for group-related traits, which is to say, suggestibility, altruism, willingness to sacrifice all, and even communal spirituality.

One clear reflection of this in human nature is the enormous importance we put on group membership versus outsider status. The ten commandments never had anything to do with non-Jews. Killing was just fine outside the group, and laws were only relevant inside the group. Indeed, their point was to cement group identity. The same is evident throughout ancient history. Greek cities of antiquity were in perpetual warfare with each other, either killing opponents or enslaving them into a life of (at best) modest reproductive success. Likewise, much of adolescence is devoted to acquiring membership and status relations within ever-shifting groups.

The lesson for today is quite simply one of globalization- the urgent need to unite humans in a single affective group, transcending the national groupings which have been so powerful both in focusing human aspirations for the last few centuries and in providing vehicles for mutual extermination. Thankfully, there are many forces carrying universal messages, from the Olympics and crass Western consumerism, to the amazing computer and cell phone revolutions that allow a global audience to follow video tweets from the streets of Iran.

Are religious messages a positive influence? Yes and no. Religious doctrine and organization has through history been perhaps the most powerful creator of group identity (and thus also enters into the group selection arguments above). The probability that any one form of religion will capture the entire market and thus unify humanity is exceedingly low (however fervently wished-for). The best hope therefore is to mitigate religious particularities, dialing down dogma such that spirituality retains an empathetic, unifying pan-humanistic strength, while the divisiveness of fabulistic and group-identifying doctrinal commitments fades away.

Incidental links


  1. Hmmm. Interesting. I agree with the basic drive of your conclusions - bring on the Star Trek age! No more money and everyone is open-minded. Prime Directive, baby.

    I wonder about labeling those who are courageous in battle as altruists. What about Vikings, Huns, Cossacks, Visigoths etc. - the violent cultures? Couldn't changes in environment be more of a factor in all this? Were the Nazis so different genetically than Quakers? I know that selection is probably more concerned with the last several hundred thousand years, more than recent history, but still, it's a point to think about. I mean, if the Nazis had developed the atom bomb first, they may have won and then how would that change successive generations - for thousands of years perhaps? And through an indeterminate "chance" that we beat them before they figured it out. I realize that selection works through indeterminate chance as well - so finding patterns to fitness seems problematic. We only have one history to view and what we define as fit is what has survived, for whatever reason.

    Evolution is obvious - but our attempts to explain traits via natural selection often rubs me more as human desire for order, than as obviously true. But I suppose it's still worth pursuing. And what I don't know about it could fill a library - but I'm learning.

  2. I enjoyed the way you assessed this paper, thanks.

    On Steven Stark's question, Bowles refers specifically to altruism as bearing a cost for actions that don't benefit you personally, or which jeopardize your inclusive fitness. Alternatively, you could think of another trait, 'Parochial Altruism' (elucidated in 'Parochial Altruism and War' Choi and Bowles (2007), also see 'Group competition, reproductive leveling and the evolution of human altruism' Bowles (2006)) in which altruism combines with disliking members of other groups.

    Also, Steve on your comment about culture, following work by Boyd & Richerson, altruism (and the coevolution of parochial altruism and war) could be explained by cultural evolution, and thus cultural transmission of particular behaviours through imitation or other learning rules. Bowles comments on this in a footnote in the paper.

  3. Hi Simon!

    Thanks for the clarification. Although I "understand" the algorithm of selection - basically that what reproduces is what continues for whatever reason (called fitness), it still seems to me that the plasticity of human psychology can trump almost anything. Of course psychology is the product of selection to, so I always feel that there is a chicken/egg loop when self-awareness and selection are considered together.

    Interesting to think about!

    Another interesting thought - the religion most committed to practicing compassion is Buddhism - but typically Buddhist monks (the most intense practitioners) do not reproduce. Yet the monks have continued for two millennia.

    Of course, the American "Shakers' did not believe in reproduction so they no longer exist.

  4. Thank you amazing blog, do you have twitter, facebook or something similar where i can follow your blog

    Sandro Heckler