Transcript MontrealEthics

Ethical Rules, Games, and Evolution
Ted Bergstrom, Economics Dept, UCSB
Our Charge for “Debate”
We know that the distinctive features of the human body,
such as our large brains, nearly hairless bodies and
dexterous hands, have evolved through natural selection
… Our social behaviour may have evolved in the same
way…
The second point of view, however, is that our social
behaviour, and the systems of ethics on which it is based,
are uniquely human, and owe nothing to the processes
that govern societies of ants or bacteria. Our bodies may
have evolved, but our ethics requires another kind of
explanation.
My Take



Evolutionary thinking has much to tell us about ethics and
the presence of altruism.
Game theory allows us to frame questions more
effectively.
Does ethics require a “another kind of explanation” from
that of the evolution of our bodies?



Well, of course, just as the evolution of our eyes require
“different “ explanations from that of our ears.
Deeper difference is “cultural evolution”. You can inherit
ethical notions from “teachers” other than your parents.
This implies different calculus of inheritance and reproduction.
Two Competing Golden Rules




``Thou shalt love thy neighbor as thyself.’’
---Old Testament: Leviticus 19:18
``Do unto others as you would have them do unto you’’
---New Testament: Luke 6:31
One rule is an exhortation to extreme sympathy, the
other to extreme symmetry.
Questions:


Are these rules different?
Why are they so extreme?
Common to many cultures

Love thy neighbor rules—Command for sympathy


Taoist version: ``Regard your neighbor's gain as your gain, and
your neighbor's loss as your loss.‘
Do unto others rules---Command for symmetry



Confucius: “Never impose on others what you would not
choose for yourself.”
Aristotle: “We should behave toward friends as we would wish
friends to behave toward us.”
Kant: Act only according to the maxim whereby you can at the
same time will that it should become a universal law.''
Hamilton’s Rule:
(A report, not an entreaty.)
 Hamilton maintains that evolutionary principles predict
that:
``The social behavior of a species evolves in such a way that
in each distinct behavior-evoking situation the individual
will seem to value his neighbors' fitness against his own
according to the coefficients of relationship appropriate
to that situation.''
Who is my neighbor? The Pharisee’s Question




What is the domain of sympathy and/or symmetry?
Old Testament, Taoists, and Aristotle seem to restrict this
domain to “neighbors” or “friends”.
Confucius, Kant, and Parable of the Good Samaritan
seem to include all persons.
Hamilton makes very specific predictions.

Individuals have sympathy only for relatives and that only
proportional to relatedness
Golden Rules and Hamilton’s Rule

When should you take an action that costs you C and
benefits another person by B?

Golden Rules: Do it if:


the person is a “neighbor” and B>C.
Hamilton’s rule: Do it if and only if;

rB>C (where r is coefficient of relatedness to recipient)
Coefficient of Relatedness


The coefficient of relatedness of two individuals is the
probability that if one has a rare mutation, so will the
other.
For sexual diploids, like ourselves, coefficient of
relatedness r is



r=1/2 for full siblings, 1/4 for half siblings, 1/8 for cousins
1/2 for parent and child, 1/4 for grandparent and child, etc.
Nearly 0 for random stranger
Are Golden Rules Unrealistic?

Believers in Homo Economicus would think so.

So would believers in Hamilton’s Rule.

Are golden rules just empty preaching?

Return to this question later.
Ethics in games




Subtleties of ethics are better understood in framework
of game theory.
Hamilton considered only a special class of “game” in
which both the cost to you and the benefit to the other
player of your own action is independent of the other
player’s action.
In this environment, the two versions of the golden rule
are equivalent.
In more general games, they are not.
An Example: A prisoners’ dilemma game
Two strategies, c and d.
 Payoff function f(x,y) is what you get if you do x and the
other person does y.
 Let f(c,c)=R, f(d,d)=P, f(d,c)=T, and f(c,d)=S, where
S<P<R<T.
 Selfish Play: Dominant strategy equilibrium is both
choose d.
 Do unto others rule. You would like other to cooperate.
So rule demands cooperate.
 Love thy neighbor rule: Choose the thing that maximizes
the sum of your payoff and other player’s.

Love-thy-Neighbor in Prisoners’ Dilemma
Love thy neighbor can lead to a trap where both defect.
 Players care equally about their own and neighbor’s
payoff.
 Suppose that T+S<2P.
 Then there is a Nash equilibrium where both defect.
 If other guy is defecting, we will both get P if I defect.
If I cooperate, he will be better off, but his gain T-P is less
than my loss, P-S.
 There is also an equilibrium where both cooperate, but
this is not unique as it is for Do-unto-others types

Love-thy-neighbor in Prisoners’ Dilemma






Prisoners’ dilemma as before.
Players care equally about their own and neighbor’s
payoff.
Suppose that T+S>2R
In equilibrium, one defects and the other cooperates.
Doing the opposite of the other guys action maximizes
sum of payoffs.
In this case, love-thy-neighbor results in higher joint
return than Do-unto-others.
Hamilton’s rule for general games.


Two possibilities:
Corresponding to Love-thy-neighbor


Corresponding to Do-unto-others




Love thy neighbor r times as well as thyself.
Act as if your payoff is H(x,y)=f(x,y)+rf(y,x)
Semi-Kantian rule: Act as if the probability is r that your
neighbor will copy you
Act as if your payoff is V(x,y)=(1-r)f(x,y)+rf(x,y)
In simple additive games considered by Hamilton, these
two rules yield same behavior.
In general, they do not.
Which Hamilton’s rule is right?





Do we expect to see evolution of “love for relatives” of
of more abstract semi-Kantian behavior?
For sexual diploids and symmetric games, the semiKantian rule is predicted by the most common model of
resistance to dominant mutant alleles.
For asymmetric role-playing games, either rule could be
appropriate, depending on the details of genetics and
cross-over.
For games with “concave payoff functions” predictions of
the two theories predict the same behavior.
Maybe love is easier to evolve.
Is Hamilton’s rule too selfish?




Why might evolution produce more altruism than
Hamilton’s rule predicts?
Common reproductive interest of partners mated for life.
Repeated interactions between any two people.
If repeated encounters mean that you will usually wind
up playing with somebody who plays as you do, then a
“semi-Kantian” preference with high r may be the most
successful under evolutionary pressure.
Had enough?
OK, I’m Done
UCSB Campus