Paradox of Animal Sociality,
Lecture #2003-11
THE PARADOX OF ANIMAL SOCIALITY: CAN A NEARLY WELL
FORMED DARWINIAN STORY BE TOLD ABOUT ALTRUISM?
A. Introduction
Some behaviors are easier to tell nearly well formed Darwinian stories about than
others. An example of a behavior difficult to tell such a story about is the helping
behavior of paper wasps.
Around the old sheds of my farm is a wasp, Polistes, who makes its living in the
summer by capturing caterpillars off my cabbage plants. Polistes pursues the following
rather odd reproductive cycle. The females are mated in the fall and winter over in
crevices in buildings. These females begin to emerge as soon as the weather gets warm
in the spring and make their way to the my sheds where each female begins to build a
nest. At that time, remember, each is a fully functional reproductive. If she continued to
build her nest and if it were to be successful, then all the offspring of that nest would be
hers.
Then a bizarre thing happens. Some of the females -- perhaps even most of them
-- leave their own nests and join other females on their nests and begin to help them raise
their broods. In so doing they give up the possibility of ever reproducing themselves. In
working on the other female's nest, they subordinate themselves, and this subordination
produces a physiological regression. For whatever reason, in only a few days, the
fertilized ovaries of these females begin to shrink and eventually these once fully
functional reproductive females become “auxiliaries” -- non-reproductive workers that
serve the reproduction of other females.
Now, we have no difficulty in seeing how this behavior is useful to the helped
female. When one female works alone at the nest, she has to leave it unguarded when
she goes off to forage for building materials and food. The nests of single females are
thus left vulnerable to predation or plundering of building materials by other species or
by other Polistes. The second female is thus a tremendous help in that she can go off and
do the foraging, while the reproductive female stays home and does the building and
guards the nest and begins to lay eggs. This arrangement is surely one that leads to the
production of more wasps. Without working together, the wasps could produce at most a
few young each. Working together, the auxiliaries give up the few young they would
have produced, but because of this sacrifice, the foundress produces dozens of young.
Overall, it seems like a very sensible arrangement.
However, figuring out how a behavior might be sensible is not the same as
figuring out how natural selection might bring it about. To explain how natural selection
might have produced and maintained a behavior, we have to show how that behavior
resulted in the production of more wasps with that behavior. There lies the problem: the
wasps, which display the behavior, don't have any offspring. They are too busy helping
to raise the offspring of wasps who don’t display the behavior! How then, has the
behavior been produced and maintained in the population?
Behavior like that of Polistes has been termed altruistic. We will spend much of
this lecture trying to get a handle on this controversial term. For the moment, we can
understand that it is behavior that provides benefits to a competitor at a cost to the to the
individual performing the behavior. Benefits, you will recall, are always reproductive
benefits.
Lets put the puzzle in a more general form. Lets imagine that in a social
organization there are Donors and Beneficiaries. Lets further imagine that Donors give to
beneficiaries some of their reproductive output. They might accomplish this generosity
in a variety of ways, some extreme, some subtle. In the extreme case, they might submit
to cannibalism. More subtly, they might not fight so hard to defend their own territory
against the territory of a fellow species member, so that their offspring ended up with less
food than their neighbors as a consequence. The mechanism is really irrelevant. All that
is necessary is that Donorship is an inherited trait and that Donors, as a group, are at a
reproductive disadvantage to beneficiaries on account of their donorship. The paradox is,
how could donorship be produced and maintained in the population?
Given that (1) a Darwinian story is one that attributes the traits of contemporary
animals to the fact that trait bearers have more offspring than non-trait bearers, and given
that (2) altruism is defined as a behavior that provides reproductive benefits that lead to
the greater reproduction of trait bearers by comparison with non-traitbearers, we can
easily see why telling nearly well founded Darwinian stories that explain altruism is
difficult. In fact the opposite is true: easily told are Darwinian stories that show that
altruism should not evolve and, if it ever did evolve, should be rapidly be replaced by
selfish behavior.
B. Darwinian stories are easy to tell that explain why altruism could
never occur.
In the previous lecture, we learned to render Darwinian stories as games tables in
which two alternative traits are pitted against one another. In this section, we develop
some models that will make clear why it is so difficult to tell well-founded Darwinian
stories concerning altruistic behavior. Here is one:
Because hens are highly social animals that have for many generations been selected
for the amount and size of the eggs they can produce, they make a useful model for
illustrating the paradox of animal sociality. Hens have strict dominance hierarchies and
making particularly large and numerous eggs presumably requires the resources that only
a particularly dominant hen can obtain. Unless hens are raised alone in individual
cages, their egg production is going to depend largely upon their aggressiveness. A hen
that can lay many eggs is a hen that can secure more than her share of the food, and such
a hen is likely to be an aggressive hen.
Consequently, when hens are selected for individual egg production, flock
productivity suffers. The reason is that, while the most aggressive chickens lay the most
eggs, the associates of the most aggressive chickens are likely to lay fewer eggs than they
would because of the interference from their more aggressive colleagues. Now, imagine
a farmer who got to wondering if he wouldn’t get more eggs from his flock as a whole if
he selected for less aggressive hens. In order to get this program started, he substitutes a
less aggressive breed for half of the chickens in his coop and then measures the
productivity of his flock as a whole. To his delight, the chickens suffer fewer injuries
and spend less time pecking each other and more time consuming mash and laying eggs.
He has fewer "star" hens that individually produce large numbers of eggs, but egg
production for the whole flock rises because the "supporting cast" is more productive.
However, not being prepared to let matters rest there, the farmer starts selecting his new
flock for individual egg-production. An evolutionary games table makes clear what
would happen.
Payoff received by
Against individuals playing
Individual hens…. ….
Nice with
probability p
Playing
Nasty with probability
(1 - p)
R
Nice
(p)(b-c)
S
p(b)
(p)(b-c)+(1-p)(-c)=
pb-pc+(-c)-(-pc)=
pb –c
(1-p)(-c)
T
Nasty
Total
Payoff
P
(1-p)( 0)
( p)(b)+(1-p)( 0)=
pb
Nice Hens will come to characterize the population ONLY if the total payoff to Nice Hens is
greater than the total payoff to Nasty ones: i.e., when pb-c > pb or when (-c) > 0: i.e., ONLY if
the cost is actually a benefit!
Being nice has both costs and benefits. The benefits might be increased time to
spend at eating and laying eggs and fewer injuries: the costs might be less access to food.
Thus, the situation bears some resemblance to the table at the end of the previous lecture
where we evaluated the circumstances under which an inherited behavior that conveyed
both costs and benefits would come to characterize the species. But unlike the situation
in the previous chapter where the costs and benefits both accrue to the trait-bearer, in this
case, the benefits accrue to both the Nice and the Nasty chickens, while the costs accrue
only to the Nice ones. (You can see this if you compare the two tables: notice that while
both costs and benefits occur in the upper left cell of the table, only costs occur in the
upper right corner where they are experienced by the Nice chickens and only benefits are
felt in the lower left cell of the table where they are experienced by Nasty chickens).
Also, if we make just a few assumptions about the sizes of the benefits and
the costs, this table becomes a cooperation dilemma table. Recall that a table is a
Cooperation Dilemma table if the cells have payoffs in the magnitudes T>R>P>S
and 2R (T+S). If we assume that benefits are greater than costs, that benefits are
positive and costs are negative, then it follows that (b) > (b-c) > 0 > (-c) and (2[b-c])
≥ ([b-c]+[b-c]), which meets the conditions for a cooperation dilemma table. So
what we have, here, is an evolutionary co-operation dilemma table.
In an evolutionary cooperation dilemma table, since the benefits accrue to both
types, they have no effect on their relative fitness and, and the outcome of the game
depends entirely on the costs. The totals column shows that the nice chickens will lose
ground in the species if there is any cost to the behavior. This rule applies no matter
how great the benefits, and no matter how trivial the cost.
The challenge to Darwinism would seem to be profound. If selection for chicken
egg production cannot generate nice chickens, how could selection for wasp egg
production ever have generated Polistes wasps that help one another to the degree that
auxiliary wasps help foundresses? To see what the problem is, lets assume that the
foundresses and auxiliaries are distinct genetic types and consider what would happen in
a population composed of 50 auxiliary types and 50 foundress types (p = (1 – p) =
50/100). Auxiliary types help other wasps at random, unless they themselves receive an
offer of help, in which case they act like foundresses. Foundress types act only as
foundresses. Let’s assume, for the purposes of argument, that the population is
neither increasing nor decreasing – thus the average reproductive rate across the
whole population is 1, including the males (i.e., a reproductive rate of 2 per female).
Assume, for the purposes of demonstration, that each auxiliary decreases its own
reproductive output by 1 (cost = -1) and increases the reproductive output of the
individual it associates with by 2 (benefit = +2. Thus, (b-c) is 1, -c is -1, b is 2, and 0,
is, of course 0. Factoring in the relative frequencies of the two types in the
population, auxiaries will receive a net benefit of 0 from their interactions with the
two other types of individuals, whereas foundresses will be receiving a net benefit of
1, as the table below demonstrates.
Given that p = 0.5,
b = +2, c = -1,
payoff received by
Individual wasps…
…against wasps playing …
auxilliary with
probability .5
foundress with
probability .5
Total
Payoff
auxiliary
(.5)(2-1)=(.5)
(1-.5)(-1)=(-.5)
pb -c=(.5)(2)-1= (0)
foundress
(.5)(2)=(1)
(1-p)( 0)
pb=(.5)(0) = (1)
Playing
Assuming that foundresses only build their own nests and auxiliaries, unless they themselves are
helped, help other wasps build nests at random, and assume that each helping auxiliary gives a
benefit of 2 and incurs a cost of -1 by helping, the nasty will increase by 1 per individual while the
nice types stand still 1.
Notice that in just one generation, the number of foundress types would increase
from 50 percent of the population to 60% (from 50/100 to 150/250) while the
auxiliary types would decrease to from 50% of the population to 40% (from 50/100
to 150/250)2. This difference would lead eventually to the elimination of the
auxiliary type. In fact, as the previous table shows, if being an auxiliary entails any
cost, it would lead to elimination of the auxiliary type. How, then, did Polistes come
to evolve this social organization? Can we ever tell a "nearly well formed"
Darwinian story that will explain such behavior? Or does Darwinism need to be
abandoned to accommodate the facts of animal social life.
C. The role of altruism in Darwinian Stories.
Behavior such as that displayed by Polistes Auxiliaries, behavior that seems to benefit the
performer of the behavior less than it benefits others, is often referred to in evolutionary
arguments as Altruistic. The basic meaning of altruism is acting against self-interest.
Thus, one class of models that are deeply entangled in the paradox is "-interest" models.
I put the hyphen before the word interest because the word is coupled with a range of
entities, such as "self-interest", "national-interest, "corporate-interest" etc. Such
interests are often subdivided. That is, amongst an individual’s "self -interests" we might
recognize "health-interests", financial-interests, social interest, etc. The core meaning that
appears to unify all these usages is that for each entity before the hyphen in an "-interest"
attribution, the attribution identifies some class of events that will make that entity "better
off" if a member of that class occurs. So, if we say that it is in the interests of Jones
health for him to drink milk every day or get a checkup every year, we are including
"drinking milk" and "getting a checkup" amongst the class of events which, if occurs,
will be good for Jones's health. Similarly, if we say that it would be in "AT&T's
financial interest to obtain Viacom" we would be implying that "obtaining Viacom was
amongst a class of events which, if it occurs, would be good for the balance sheet of
1
In case the numbers are bothering you, as well they might, here is how they work. The population
started out with 50 “foundress-type” wasps and 50 “auxiliary-type” wasps. Assuming that the
population was in balance before the auxiliary type was introduced, females working alone (as
foundress types still do) have to be having two offspring, one for herself and one for the male who
mated her. For auxiliaries (“cooperators”) their reproductive rate stays at 2, and their total output
is therefore 100 wasps; for foundress-types (“defectors”) overall, the rate increases to 3 and their
total output is 150 wasps. Thus at the end of one generation there will be 150 foundress-types(60%)
and 100 (40%) auxiliary types.
AT&T. Also implicit in interest attributions is a scale of value by which effects are
evaluated. The presence of such scales of value in the argument is often not evident
because they are so obvious. For instance, we all assume that a positive effect on the
balance sheet of AT&T would be an increase in the amount or dollars represented there.
Even in ordinary language, there is an ambiguity in interest models that can lead
to confusion. If we say that it is in Bill Gates's financial interest to obtain Apple, the
Bill Gates we imagine after the obtaining is the same Bill Gates we imagine before it.
The same would probably true of our imaginings of the proposition that it is in Steve Jobs
interest to obtain Microsoft or Microsoft's interest to obtain Apple. However, what of the
proposition that it is Apple's interest to obtain Microsoft. Given how such mergers occur
and the relative size of the two corporations, could we really imagine that the Apple
Corporation that had obtained Microsoft would be the same after the merger? Could we
really imagine that, after such a merger, Apple would continue to exist? Thus, the
interest starts to break down when we are no longer sure that the entity that has the
proposed interest would be the same entity after the interest had been actualized.
The same ambiguity concerning interest models exist in biology. Some such
attributions seem uncontroversial. When we assert that it is in the interest of the hawk to
catch the mouse (or in the interest of the mouse to escape the hawk), we assert that the
hawk that eats the mouse or the mouse that escapes the hawk will be the better for it, but
not so much better as to become a different creature. No problems here. However,
other interest attributions in biology are much more troublesome. What, for instance, do
we imply when we say that it is in the interest of the hawk or the mouse to breed and
have offspring? The individual hawk, after breeding and mating, will not be the better
for it. In fact, the breeding process is usually dangerous and debilitating. What's more,
because of genetic recombination, the products of a breeding process are not the same
entity, any more than an Apple that merged with a Microsoft would be the same Apple.
So, when we are talking about an animal's mating and breeding, we are talking about a
different sort of interest than when we are talking about its having a good meal and a
good night's sleep.
Whatever sorts of interest this latter type is, it is the sort of interest that ought to
concern a Darwinian. In Darwinism, the essential creative force is the differential
reproduction of entities. Surely, therefore, the interest the organism has in courting,
mating, and raising young, its "reproductive-interest", if you will, is the only interest that
counts. Yet, because of the facts of reproduction, the entity produced is not the same
entity as that to which we input the interest.
One solution to this difficulty would be to descend to the genetic level and speak
of the interest of the genes for various traits. This is, of course, Dawkins's solution.
Thus, altruism would be redefined as acting against the collective interest of one's genes.
We might, for instance, say that it is in the interest of the hawk's hunting genes for the
hawk to raise offspring, meaning that if the hawk died without bearing young, there
would be no such genes. This way of speaking doesn’t raise the difficulty of a change in
the entity that has an interest actualized because the "gene for" hunting in the hawk-
parent is the same as the gene for hunting in those of its offspring that receive the gene
and the offspring-rearing of the hawk helps to spread the hunting gene. To return to our
corporate interest model, for moment, this solution would be like saying that a merger
between Apple and Microsoft might be in the interest of Apple's stockholders, even
though Apple would cease to exist as an entity. Apple stockholders receive Microsoft
stock as part of the merger and they might well earn more money on their Microsoft stock
than they had been on their Apple stock.
The metaphor seems at first to be promising. Like genes, stockholders are elements of
corporations that survive important changes in the corporation. However, to make the
metaphor work, we would have to overcome some obvious disanalogies. For
stockholders to be like genes, each would have to control some feature of the
corporation's structure; say, one stock holder might control the letter head, another type
the choice of car in the corporate car fleet, a third the choice of airline flown by the
corporations executives when they travel, and so forth. Some stockholders might have a
hand in many decisions; others might collaborate to make a single decision. But each
stockholder would have a competence that would affect the structure and behavior of the
corporation in one or more particular ways.
To make the corporate analogy work, we would also have to reconceptualize a
merger. During a merger, each stockholder at Apple that controlled a given feature of
the corporation would meet with his equal number at Microsoft, clone themselves a
bunch of times, and then they would get distributed to a bunch of new corporations
containing different combinations of Microsoft stockholders and the features they control
with apple stock holders and the features they control. Thus, for instance, some
companies might end up with Microsoft letterhead and Apple automobiles while other of
these offspring companies might end up with Apple letterhead and Microsoft fleet of
automobiles. These derivative companies would compete and the most productive
would get to go on to the next merger.
Now, if we modify the corporate model in this way, what do we conclude about
the statement that it would be in the stockholders' interest for Apple to merge with
Microsoft? Well, the answer is, not in the interest of ALL stockholders. I think some
stockholders would profit from the merger and some would not. Thus, there would be
no single "stockholders" interest that the behavior of the corporation represents.
Similarly, there is no single "genetic" interest that the behavior of the gene-bearer
represents and altruism, therefore, cannot be understood as behavior that opposes -- or
advances -- the interests of genes.
For all of these reasons, I conclude that because the concept of "blank-altruism" is
rooted in the concept of "blank-interest", it is a concept that is not going to bear much
weight. My policy, here, therefore will be to use the concept warily. In the context of
this Darwinian argument, an entity (E) will be understood to have a Darwinian E-interest
when I can imagine a state of affairs that would the reproduction of that entity. Recall
that in this context a benefit is always an increase in reproduction over the competition.
The entity could possibly be a gene, an individual, a group or species. I will measure the
success at reproduction by counting the relative number of offspring, propogules, copies
etc, of that entity that themselves reach reproductive age3. Any event that increases this
number will be referred to as a benefit to E, any event that decreases this number, a cost.
With this conceptual equipment in place, I can now make a distinction between
Darwinian Selfishness and Darwinian Altruism. I will call a trait selfish if is hereditable
and governs the actualization of a Darwinian interest in the individual that bears the trait.
I will call a trait altruistic when it actualizes the interest of one or more a competing
entities and an entity a Darwinian altruist if it has a hereditable trait that governs the
actualization of an interest for an entity other than the entity in question. The entity will
be judged an altruist whether or not the receiver of the benefit is a relative or otherwise a
bearer of the altruist trait. In fact, on this definition, the only other entities to which
benefits can be conveyed without altruism are offspring.
Our focus on Darwinian interests does not, of course, preclude the possibility that
there are other kinds of interests. For instance, we might speak of the somatic interest of
an individual, referring to all the events that might occur or all the things the individual
might do that would make that individual fatter, healthier, happier, etc. Very often, of
course, somatic interests are in accord with Darwinian ones. But not always. You may
have noticed that raising teenaged children (or putting them through college, for that
matter) has not seemed to have made your parents fatter, healthier or happier, even
though sociological and economic analysis would seem to suggest that it will increase the
number of their direct descendants.
The paradox of animal sociality may now be restated as follows. If Darwinian
natural selection favors only selfish entities (i.e., entities with selfish traits, or entities that
actualize Darwinian interests), why is it that so much behavior is (or appears to be)
altruistic?
3
OK. I can hear my readers screaming, what about a trait that made sacrifices in the number of
children to favor the number of grandchildren. Wouldn’t that be in the Darwinian interest of the
trait-bearer. I agree. So a Darwinian interest actually refers iteratively to the number of
descendants in the direct pedigree of the traitbearer. For most purposes, this number is fairly
represented by the number of offspring. Not always, though, as we shall seen in the next lecture.