The Coevolution of Altruism, Parochialism and War - Santa Fe Institute

The Coevolution of Altruism, Parochialism and War:
Rambo meets Mother Teresa
Samuel Bowles, Santa Fe Institute & University of Siena
Rock painting, Drakensberg Mts. S. Africa

In previous episodes...
• We are a uniquely cooperative species, joining with large
numbers including non-kin in the pursuit of projects for
mutual benefit
• Important aspects of cooperation cannot be explained by
self interst with a long time horizon or other extensions of
the ‘somebody may be looking paradigm.”
• Part of the explanation is that many of us are altruistic much
of the time.
• Today: how did we get to be this way?
• Based in papers in Science 314 (2006) and Science 318
(2008) Nature (in press) and Journal of Theoretical Biology
(2003), and on-going unpublished work

Reproductive levelling
Ethnic boundaries
Paleoeconomics
Human nature
War
Economic man
Cooperation
Self interest
Altruism

Fights, Food and Fitness

Peter Kropotkin
...mutual
aid..
..red in
tooth and
claw
Human nature and war
Thomas Huxley
...Selfish and contentious people will not cohere, and
without coherence, nothing can be effected. A tribe
possessing ... 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...Thus the social and moral qualities
would tend slowly to advance and be diffused
throughout the world. Darwin, Descent of Man, 1873
Do we engage in mutual aid because evolution is red in tooth and claw?

Do we engage in mutual aid because evolution is red in
tooth and claw?
?

Darwin’s idea, modernized
• 0. Groups. We randomly assign n
individuals to g groups. At t=0 all
are N;
• 1. Pairing. In each period each
member of a group is randomly
paired to play the PD game once,
with another member.
• 2. Reproduction. Replicas of the
current generation constitute the
next gneration. They are produced
by drawing (with replacement) from
the current group membership with
the probability that any member
will be drawn equal to that
member's share of the total payoffs
of the group.
• 3. Mutation. With probability e a
member of the next generation is
not a replica of its parent, but is A
or N with equal probability

• 4. Migration. With probability m
each member of the new generation
relocates to a group randomly
selected from the other groups.
• 5. Group competition. With
probability k each group is selected
and among those selected
competition takes place among
randomly paired groups. The
winning group is that with the
highest total payoff (net of the costs
of sharing and segmentation if any.)
• 6. Repopulation and fission. The
members of the losing group are
replaced by replicas of the members
of the winning group, and the
resulting (temporarily enlarged)
winning group splits with members
assigned randomly to two new
groups.
Sounds like a plausible story if you haven’t studied biology…

Warning: group selection alert!
• Darwin’s account is termed group selection. And 40 years
ago biologists put it in the dog house of bad ideas
• Unworkable because for genetically transmitted traits, it
requires that
– the ‘victorious tribes’ would have to be very different genetically
from the losers, and
– such encounters among tribes would have to be frequent and lethal.
• The first was thought not to be true due to migration among
groups; and
• ..the second encountered resistance from an idyllic
conception of the world before states.

Why biologists thought that altruism
beyond the immediate family could not
evolve: The prisoners dilemma
A N
A b-c -c
N b 0
•An altruistic behavior costs the individual c and confers a
benefit of b on a randomly paired (single) member of the
group, so a member in a group composed entirely of
altruists produces b-c > 0 more replicas than that of a
member a group with no altruists.
•This is a prisoners’ dilemma (N has higher fitness no matter
who it one is paired with).
•Warning: short detour through some heavy lifting. Don’t
try the next few slides at home!

Evolutionary
fate of an
altruistic trait
A
N
A b-c -c
N b 0
NB: P(A|A), P(A|N) are
respectively the probability of
being paired with an A given that
one is an A or an N
•With random pairing, payoffs to the As = π A < π N = payoffs
to the Ns for every frequency, p, of As, dooming the A’s
• But if groups differ in their frequency of As, the As benefit
from more frequent interaction with As i.e P(A|A) > p >
P(A|N ) so the doom scenario may not hold.
• P(A|A) - P(A|N) ie. how different groups must be in order
for the As to survive

Out of the dog house?
• Surprising recent evidence suggests that many of our
foraging ancestors lived in genetically differentiated and
warlike groups, and adopted other practices likely to
make Darwin’s account work.
• Amassing and marshalling this evidence in a coherent
analytical framework required population genetics,
anthropology, archaeology, even climatology, using
techniques as varied as interpretation of skeletal remains,
analysis of genetic markers, computer simulations,
ethnography, dynamical systems analysis, and game
theory.
• I begin with the evolutionary modeling

Why humans
escaped the
evolutionary fate
of an altruistic trait
A N
A b-c -c
N b 0
• The degree of positive assortment is P(A|A) - P(A|N)
• How different groups must be in order for the As to survive
shown in the figure = c/b namely, the ratio of the individual
costs to the group benefits.
•Those who put group selection in the dog house believed
that between group genetic differences, namely P(A|A) -
P(A|N) were much too small small, maybe on the order of
0.02 (requiring b to be 50 times c!)
• Others also thought that b/c would be quite small (2, 4, etc)

• The outcome is a horse race between the within group
selection processes leading to the elimination of A’s
within each group and the between group selection
processes benefiting the A’s.
• Most biologists have thought that the within group
horse would win

Human distinctiveness to the
rescue!
• But humans, maybe uniquely among animals, proved
very good at raising both the group benefits of altruism
(b) and genetic and other differences between groups
(P(A|A) - P(A|N)) and lowering the costs of being an
altruist (c ), giving Darwin’s idea a chance.
• The good news: leveling – monogamy, sharing food, --
lowered c
• The bad news: hostility towards outsiders and frequent
wars raised b and P(A|A) - P(A|N)

An example: Reproductive or material leveling institutions
such as food sharing among non-kin and monogamy reduce
within-group differences in reproductive or material success.
• By reducing payoff differences within groups, such leveling
structures reduce c and attenuate within-group selective
pressures against the As, favoring the survival of the groups
adopting them.
• So both the As and the leveling that protects them survive!
• Leveling practices are common among recent foragers and
were almost certainly so among our forager ancestors.
• In this case the evolutionary success of group structural
characteristics such as leveling institutions is explained by
their contribution to the proliferation of group beneficial
individual traits.
• Leveling institutions create a niche favorable to the A’s

Resource Sharing (Tax) on Evolution of Altruism
Adding segmentation
Tax
π N
π N
π A
b
b-c
(1-s)b b-ct
P(A|N)
½
P(A|A)
π A
b-c
p
sb-c -(1-t)c
P(A|N)
½
P(A|A)
p
Note: -c
the payoffs to the A-trait are less than the payoffs to the N trait for every
Frequency, p, of the A trait.
Institutions allow A to evolve for smaller P(A|A)-P(A|N) (i.e.
smaller between group genetic differences)

Similar ideas have been suggested for human societies
(Alexander 1979, Boehm 1982, and Eibl-Eibesfeld 1982), and
non-human organisms (Keller 1999, Maynard Smith and
Szathmary 1995, Michod 1999: Ratnieks, 1988, and others)
Frank, Steven A., “Mutual Policing and Repression of Competition
in the Evolution of Cooperative Groups,” Nature, 377 (12 October,
1995): 520-522.
“Evolutionary theory has not explained how competition
among lower-level units is suppressed in the formation of
higher-level evolutionary units,…mutual policing and
enforcement of reproductive fairness are also required for
the evolution of increasing social complexity.”
Nota bene, economists: suppression of competition (not
competition itself) is the key to evolutionary success in this
case!

Co-evolution of individual behavior and group level
institutions
• Remember for A to succeed differences between
groups must exceed the ratio of costs to benefits, or
P(A|A) - P(A|N) > c/b
• The synergy between leveling and altruism lowered c,
the costs of altruism; but theres more to the story.
• Frequent warfare raised b
• And the hostility towards outsiders both raised b and
(by reducing migration) increased P(A|A) - P(A|N)

Why agent based computer simulations?
• We are asking a hard question: why something tht we
know happened in the distant past occurred (a causal
statement, not just descriptive).
• The process we are investigating is way too complex to
allow illumination by mathematical modeling (it can be
modeled, but one cannot extract answers from the results).
• Like experiments, agent based modeling allows us to
independently vary influences on behavior while holding
constant others.
• In this case we vary such things as migration between
groups, group size, differential mortality among winners
and losers, mating practices and the like.
• For starts: the effect of varying the frequency of war…

1
Fraction of As
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
War and leveling
resource sharing allow ed resource sharing not allow ed
0
1 4 7 10 13 16 19 22 25 28 31 34
Probability of Group Conflicts (k)
• Each data point is the average frequency of altruists in the
entire population over 10 runs of 50,000 periods
• Frequent war facilitates the evolution of altruism
• The same for limited migration, small group size
• Where groups can adopt resource sharing (leveling
practices), groups that adopt these tend to survive longer
(sharing about 20% of food).. and …
• a given amount of warfare supports a higher fraction of A’s

1
Fraction of As
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
War and leveling
resource sharing resource sharing not
1 4 7 10 13 16 19 22 25 28 31 34
Probability of Group Conflicts (k)
• The model shows that a given amount of warfare supports a
higher fraction of A’s;
• But that’s cheating because we have just assumed war rather
than explained it.
• We need to expand on Darwin’s idea

Darwin amended: Do we engage in mutual aid because
evolution is red in tooth and claw? And is our evolution red
in tooth and claw because we engage in mutual aid?
?
If wars are frequent and altruistic groups win, then the
answer to the first question is yes. But why are humans
such a warlike species?

Parochial altruism: the problem
• Parochialism --favoring ethnic, racial or other insiders over
outsiders – and altruism is commonly observed human
behaviors that are well documented in experiments (insider
favoritism is far from universal, however).
• Parochials forego beneficial interactions with ‘outsiders’
e.g. exchange, co-insurance, and access to information
• Parochial altruism is puzzling from an evolutionary
perspective because both reduce the actor’s payoffs
(whether fitness or material well-being) by comparison to
other group members who eschew these behaviors.
• The big idea: neither parochialism nor altruism is could
evolve separately, but the synergies between them allow
for their co-evolution.

Consider a population in which
P T
individuals may be either Altruistic (A)
or Not (N) and either Tolerant (T) or A PA TA
Parochial (P) towards other groups (these N NP NT
are 4 behaviors, not preferences)
• A’s contribute to the fitness of other group members at a
cost to themselves
• Only the PA’s fight wars.
• P’s induce hostilities and forgo the benefits of peaceful
interactions with other groups enjoyed by the T’s
• In every group Tolerant beats Parochial and Not beats
Altruist
• So (the key point): NT is the dominant strategy, the other 3
types will be eliminated; both P’s and A’s are doomed.
• War to the rescue!

Four behavioral types; two
selection processes
P T
A PA TA
N NP NT
Within group: public goods
game with benefits b shared
equally among n members
and costs c to the contributor;
NT dominant strategy
Between groups: (a) hostile
conflict or (b) trade, insurance,
exploiting buffer zones: may
favor PAs
Behaviors are transmitted vertically from parents with higher
payoff parents making more copies (transmission may be
genetic or cultural).

Does the model describe our ancestral past?
“…towards the end of the
Pleistocene as anatomically
modern humans began to
emerge, group extinction rates
could have risen dramatically
as needy bands of well armed
hunters, strangers lacking
established patterns of political
interaction frequently collided,
either locally or in the course
of long distance migration.”
Christopher Boehm
Healed previous arm (left ulna) fracture

Empirical plausibility: Pleistocene and Holocene Temperature Variation
Oxygen isotope signatures from ice cores from Greenland (17,496
observations) Surface temperature scales approximately linearly with the
δ 18 O. Differences in (C) temperature are about 1.2 times the difference in
the signal shown the figure.

Genetic differentiation & mortality in warfare among foragers:
sources of data.
■ indicates archaeological sites ● refer to ethnographic data, and lines
point to the areas over which genetic differentiation data were collected.

Fraction of deaths due to intergroup conflicts (δ) in hunter
gatherer populations : ethnographic evidence
Population
(region)
Ache, (Eastern
Paraguay)
Hiwi,
Venezuela-
Colombia
Murngin
NE Australia:
Ayoreo Bolivia-
Paraguay
Tiwi
N.Australia
Modoc
N. California
Yuki
N.California
Kato
N.California
Yurok
N.California
Dates
Livelihood and
society
Citation
pre-contact (1970) foragers (44) Hill and Hurtado
(1996)
δ
0.37
pre- contact (1960) foragers (45) Hill et al. (2007) 0.17
1910-1930 forager including
maritime
1920-1979 seasonal foragerhorticulturalists
1893-1903 sedentary hunter
gatherers
1934 field work re
‘aboriginal times.’
before 1850
before 1850
sedentary huntergatherer
sedentary huntergatherer
sedentary huntergatherer
1830-1840 sedentary huntergatherer
(storage)
(46) Warner (1931) 0.23
(47) Bugos (1985) 0.15
(48) Pilling (1968) 0.02
(49) Ray (1963) 0.13
(50) Kroeber (1953) 0.27
(50) Kroeber (1953) 0.01
(50) Kroeber (1953) 0.04

War in hunter gatherer society: archeological evidence
• Near Wadi Halfa in the Sudan 58 skeletons dating from
12-14k ybp were found along with 189 flaked stone points
and barbs of spears or projectiles, many of which were
lodged in the vertebral column, chest cavity and skull . The
deceased had been large savannah mammal hunters and
occasional fishers.
• The archeologist who excavated the site (Wendorf)
remarked:Violence must have been a very common event in
Nubia at this time, if we are to consider this graveyard as
typical. There appears to be no significant distinction
between males, females and children in their exposure to
violent death; evidently all members of the group were
involved in conflict, not just the adult males.

19 stone points were
embedded in or
associated with the
(former) individual
on the left (Wadi
Halfa burial, 12-14k
ybp)
A study of arrow
wounds treated by U.S.
Army surgeons during
the Indian Wars found
that less than a third of
the arrows struck bone
(Milner (2005)).

Archeological evidence: δ = % of deaths due
to violence among hunter gatherers
Site, source Date Citation δ
N. British Columbia 3500BCE-1774AD (37) Cybulski (1994) 0.22
Nubia (site 117) 12-10000 BCE (26) Wendoff (1968) 0.46
Nubia (Qadan) 12-10000 BCE (26) Wendoff (1968) 0.03
Ukraine (Vasylivka) Mesolithic (38) Vencl (1991) 0.18
Ukraine (Volos’ke Mesolithic (38) Vencl (1991) 0.11
S. California 3500BCE-1380AD (25) Lambert (1997) 0.07
Central California 1500BCE-500AD (39) Moratto (1984) ¥ 0.05
Denmark (Vedbaek) 4100BCE (40) Price (1985) 0.14
Sweden (Skateholm I) 4300BCE (40) Price (1985) ¥ 0.038
A piercing wound in the
left innominate (hip)
Southern California
(Lambert (1997)
Central Ca (SCL 674) 415BC-235AD (41) Andrushko et al (2005) 0.08
Central Ca (SCI-038) 240BC-1770AD (42) Jurmain (2001) 0.04
Central Ca (Ala-329) 500AD – 1770AD (42) Jurmain (2001) 0.04
Gobero, Niger, 14,000–6200 BCE (43) Sereno et al.( 2008) 0.00

Were prehistoric forager groups genetically different enough
so that group selection would have been a powerful
evolutionary force: Evidence from ethnographic foragers (the
average is 4 times the previous consensus view)
Source: Bowles, Science 2006

Were goups warlike enough and
different enough for Darwin’s
idea to work?
• Taking median estimates of
genetic differences and wartime
mortality, if groups with more
altruists were proportionally
more likely to survive, quite
costly forms of altruism could
have proliferated.
• But we still have to explain why
warfare was this common.
• To do the we play Artificial
History

Between-group interaction game tree:
frequent interactions may favor APs
f AP , f P fraction AP, P, etc
) = difference in number
of ‘fighters’
Groups are repopulated
by couples randomly
formed from the
survivors in the winning
group

Altruists
Parochial Altruists
Parochials
Prob (war) per generation per group
The co-evolution of altruism, parochialism, and war
Shown: transitions from selfish peace to altruistic war
(and back)

The height of the bars
gives the fraction of a
very long period in
which we observe the
indicated pair of
population level
frequencies of altruists
and parochials in the
population.
An empirically estimated stationary
(ergodic) distribution
Try me!

Conclusion:
• Under conditions approximating those
experienced by our Late Pleistocene
ancestors, groups of parochial altruists could
emerge, and such groups would frequently
engage in and win hostile conflicts with
other groups.
• Independent of initial conditions, neither
parochialism nor altruism is viable singly
but warfare, altruism and parochialism
could have evolved jointly.
• Cooperation against others may be the
distinctive capacity of our ancestors that
explains their success in the great exodus
from Africa around 60 thousand years ago.

Our legacy is not our destiny
• No evidence of a genetic basis for war,
altruism or parochialism; what we show is
that if alleles supporting these behaviors
were to exist they might have evolved.
• Parochial altruism may be our (genetic and
cultural) legacy, but it need not be our fate.
• This uniquely cultural species can invent,
redirect, and overcome, outsider hostility
through socialization exposure, etc
• Examples: experiments, political
movements, and a story.
Rosa Parks

Where does this leave us, as a species?
• We are a uniquely cooperative species.
• Our cooperative nature cannot be explained by self interst
with a long time horizon or other extensions of the
‘somebody may be looking paradigm.”
• Part of the explanation is that many of us are altruistic much
of the time.
• Both levelling and warfare played an essential part in how
we got this way.
• Tomorrow: how can this knowledge improve the way we
govern our local, national and global interactions to provide
a flourishing and sustainable life for all humans.
• Warning; no easy answers. A day care center in Haifa will
start us off.

Thanks to my co-authors Herbert Gintis,
Astrid Hopfensitz, Jung Kyoo Choi
and Sung Ha Hwang, and to Margaret
Alexander, Tim Taylor, Della
Ulibarri, Bae Smith, and the other
staff of the Santa Fe Institute, and to
my countless critics for their
contributions to this research, to the
Cowan Endowment for the Behavioral
Sciences Program at the Santa Fe
Institute, the NSF, the European
Science Foundation, and the
University of Siena for support of this
research
Related papers at http://www.santafe.edu/~bowles