CEFOM/21 Third International Symposium

1
CEFOM/21 Third International Symposium Trust in
Groups from Cross-Societal Perspectives Hokkaido
University, September 26-28, 2003 An
Institutional Approach to Groups
Groups are important because people think that
groups are important,
but, one person alone cannot change her beliefs
about the importance of the group.
Toshio Yamagishi Hokkaido University
2
We intuitively think that groups play an
important, indispensable role in our social life.
Why are groups so important? Our intuitive
answer to this question is that groups are where
cooperation takes place. We are better off as a
member of a group than by living independently.
Why does cooperation take place mostly within
groups?
We do not have an answer to this question! Most
answers provided by psychologists are too
intuitive and weak in logic. Game theorists have
not come up with a satisfactory model of ingroup
cooperation.
3
Why does the group promote cooperation?
A psychological mechanism that makes us behave
cooperatively toward ingroup members is not
acceptable as an answer to this question, since
it immediately prompts another question
Why do we have a psychological mechanism that
makes us behave cooperatively toward ingroup
members?
Psychologists approach to this question is to
analyze the mechanism itself This approach
answers the question of how, but does not answer
the question of why.
The game theoretic approach answers the question
of why, though it does not provide clues to the
question of how.
We need to combine the two approaches to come up
with an acceptable answer to this question!
4
Why does the group promote cooperation? What do
we know about why we cooperate?
Cooperation Behavior that promotes another
persons (or other persons) welfare with a
non-zero cost to the cooperator him/herself
(where C lt B). Altruism no constraint of C lt B
? kin-selection in evolutionary biology is
relevant to our question only to the extent that
groups consist of kin members. Since our question
is about the group in general, and not limited to
kin-based groups, kin-selection does not provide
an adequate answer to our question, unless our
psychology confuses cues for the group with the
cues for the kin. ? Adaptive advantage of
helping kin members quickly diminishes beyond
cousins. The psychological mechanism that
confuses non-kin group members with kin members
is fitness reducing unless groups consist mostly
of close kin members.
5
Why does the group promote cooperation? What do
we know about why we cooperate?

1. Reciprocal altruism (evolutionary biology)
   tit-for-tat strategy (game theory)

TFT-based mutual cooperation in iterated PD
involves direct exchanges between particular
partners. There is no logical (game theoretic)
reason that groups matter.
Even if empirical research shows that PD players
cooperate more with ingroup members than outgroup
members, TFT-based cooperation cannot explain the
fact.
Reciprocal altruism or TFT does not provide an
adequate answer to our question.
6
Why does the group promote cooperation? What do
we know about why we cooperate?

• Reputation-based generalized exchanges (giving
  game)

Each actor decides whether to give resources or
not to another actor. (C lt B) Actors who give
earn a good reputation. A strategy to give only
to others who have a good reputation can evolve
several game theoretic models are currently
competing.
7
Why does the group promote cooperation? What do
we know about why we cooperate? ?Reputation-based
generalized exchanges (giving game)
Each actor decides whether to give resources or
not to another actor. (C lt B) Actors who give
earn a good reputation. A strategy to give only
to others who have a good reputation can evolve
several game theoretic models are currently
competing.
Unlike TFT, this strategyimage scoring,
standing, etc.does not require shadow of the
future (high probability of future interactions
with a particular other).
Insofar as the reputation information is
transparent to everyone, and all actors regard an
actor who gives her resources to someone else
(regardless of which group the target belongs) as
good, then the group plays no role in the
promotion of cooperation.
8
Insofar as the reputation information is
transparent to everyone, and all actors regard an
actor who gives her resources to someone else
(regardless of which group the target belongs) as
good, then the group plays no role in the
promotion of cooperation.
Reputation-based cooperation is hard to maintain
when some of those who may potentially give
resources to an actor does not have an access to
the actors reputation (i.e., when only a
fraction of people who potentially give to an
actor knows her reputation). RBC is viable when
Most of the potential givers to an actor share
the actors reputation
Cooperation is maintained only within groups if
we defined the group as a collection of people in
which they share each others reputation and have
a chance to give to everyone else.
9
We thus can explain cooperation within groups,
based on the RBC model, if we define the group as
a community in which everyones reputation is
shared by all and all have a potential to help
the others. What matters is whether a particular
group one finds herself in is a community in this
sense.
Group identity or a shared marker does not play
any role in this explanation. What matters is the
existence of the community a marker is not
needed to define such a community. Does the
shared marker play any role in the maintenance of
cooperation among those who share the marker?
10
Does the shared marker play any role in the
maintenance of cooperation among those who share
the marker?
To answer this question, let us think of a
population in which all can potentially help all
others and all actors reputations are shared by
all. Furthermore, the whole population is divided
into groups by a marker in this example, a
color.
Does the marker affect RBC in this population?
No!
unless,
actors believe that markers are important.
11
Does the shared marker play any role in the
maintenance of cooperation among those who share
the marker?
A marker matters when and only when actors
believe that others use markers in defining who
have a good reputation. Beliefs about
how a good reputation is generated.
If actors believe that giving to their own kind
(who share the same marker) is considered good
by their own kind (and those who dont share the
marker are indifferent to the behavior), then the
marker plays a critical role in the actors
decision of to whom to give.
12
Does the shared marker play any role in the
maintenance of cooperation among those who share
the marker?
But, where does the belief come from? The
belief-based ingroup cooperation is a tautology
unless we can answer this question. The answer
should be sought in the adaptive advantage of
having such a belief (and adjusting behavior
according to the belief). ?A belief is
self-sustaining when having that belief is
adaptively advantageous. ?A self-sustaining
system of shared beliefs institution (Aoki,
2003)
?The belief-based ingroup cooperation exists only
as an institution!
13
An institution as a self-sustaining system of
shared beliefs
Social Reality An incentive structure in which
having a particular set of beliefs and behaving
according to the beliefs is adaptive
Beliefs Actors acquire the set of beliefs and
behaving according to the beliefs, since doing so
is adaptive.
Beliefs are not necessarily reflections of the
social reality! What matters is the adaptive
advantage of having the belief.
14
An institution as a self-sustaining system of
shared beliefs
Cultural Psychological Conception of
Self-perpetuating system of beliefs is similar to
the conception of an institution, but it lacks
the logical foundations.
This may be true, but it is useless in the
logical analysis of the role of the group in
promoting cooperation.
15
An institution as a self-sustaining system of
shared beliefs
The Naïve Theory of Groups
Adaptive advantage of sharing the beliefs Having
the naïve theory of groups protects one from
making an error of disregarding group boundaries.
Social Reality Those who ignore group boundaries
and treat both ingroup and outgroup members in
the same manner acquire a bad reputation (the
ones who dont help us) and fail to be treated
favorably with either ingroup or outgroup members
Beliefs The naïve theory of groups Groups are
where people help each other (including
generalized exchange)
Those who share the beliefs engage in
ingroup-favoring behavior
16
An institution as a self-sustaining system of
shared beliefs
Insofar as the majority of people in a society
share the naïve theory of groups (collectivistic
theory), it is advantageous to believe in the
theory. It protects people from committing costly
errors. The naïve theory sustains itself in a
self-fulfilling manner.
Insofar as people share the belief that most
people treat cooperation with or altruism toward
ingroup members as good, it is better to behave
in ingroup-favoring manners. That behavior
reconfirms the belief in a self-fulfilling
manner. What matters is the expectation about
other peoples beliefs ? meta belief belief
about other peoples beliefs Our experiments have
demonstrated that the belief that others behave
in an ingroup-favoring manner is behind high
cooperation with ingroup members,
ingroup-favoring reward allocations, and trust in
ingroup members.
17
Experimental evidences that the belief that
others behave in an ingroup-favoring manner is
behind the high cooperation with ingroup members,
ingroup-favoring reward allocations, and trust in
ingroup members.
Example 1 Ingroup-favoring reward allocation in
the minimal group experiment (Tajfel et al., 1972)
Minimal Group Experiment (Tajfel et al.,
1971) Ss are divided into two group on a trivial
criterion. Preference of Klee and Kandinsky,
estimation of the number of dots on a slide,
etc. There is no interaction within group
members, and there is no conflict of interest
between the groups. Each S allocates rewards
(money) between one in-group member and one
out-group member. Ss in such minimal group
experiments often allocated more money to an
in-group member than to an out-group member. ?
Social categorization alone produces in-group
favoring behavior.
18
The belief that others behave in an
ingroup-favoring manner is behind the
ingroup-favoring reward allocation in the minimal
group experiment (Karp et al., 1993)
Money allocated to an ingroup and an outgroup
member (out of 500 yen) by Ss who were also a
target by other subjects' allocations (Target
Condition) and those who were paid a fixed amount
(Not a Target Condition)
19
The belief that others behave in an
ingroup-favoring manner is behind the
ingroup-favoring reward allocation in the minimal
group experiment (Karp et al., 1993)
The same result was replicated by Jin, Yamagishi
Kiyonari (1996)
20
The belief that others behave in an
ingroup-favoring manner is behind the
ingroup-favoring reward allocation in the minimal
group experiment (Karp et al., 1993)
Furthermore, only those who expected their favors
to the ingroup member to be reciprocated
practiced ingroup favoritism.
21
The belief that others behave in an
ingroup-favoring manner is behind the high level
of cooperation with ingroup members (minimal
groups) (Jin Yamagishi, 1997)
Cooperation in one-shot PD with an in-group
partner and an out-group partner (Cooperation
rate proportion of the endowment of 100 yen
given to the PD partner).
Ingroup favoring cooperation typically observed
in earlier experiments was replicated in this
experiment Ss were more cooperative when they
played with an ingroup member than with an
outgroup member.
22
The belief that others behave in an
ingroup-favoring manner is behind the high level
of cooperation with ingroup members (minimal
groups) (Jin Yamagishi, 1997)
However, ingroup favoring behavior did not exist
when the ingroup partner did not know that the
subject was in the same group.
23
The belief that others behave in an
ingroup-favoring manner is behind the high level
of cooperation with ingroup members (minimal
groups) (Jin Yamagishi, 1997)
The same pattern was obtained by Kiyonari (2000)
more clearly.
Whether the partner was an ingroup or an outgroup
member did not matter unless the player could
expect ingroup-favoring behavior from other
in-group members.
24
The belief that others behave in an
ingroup-favoring manner is behind the high level
of cooperation with ingroup members
Nationality (Japanese versus Australian), not the
minimal group
Japanese Data IngroupJapanese OutgroupAustralian
Australian Data IngroupAustralian
OutgroupJapanese
25
The belief that others behave in an
ingroup-favoring manner is behind the group-based
trust (Foddy, Platow Yamagishi)
Ss (students of La Trobe University) had a choice
between two dictators, one from the same school
and the other from Melbourne University. Each
dictator was given AUD16.00 to divide freely
between him/herself and a recipient. In the
unilateral knowledge condition, Ss (recipients)
knew the group membership of the dictator but
the dictator did not know the recipients group
identity. In the mutual knowledge condition, both
the dictator and the recipient (subject) knew
their group identities.
26
The belief that others behave in an
ingroup-favoring manner is behind the group-based
trust (Foddy, Platow Yamagishi)
Percentage of the choice for ingroup allocators
Trust in the ingroup member is based on the
belief that ingroup members behavior is guided
by the naïve theory of groups!
27
Remaining Issues
Do people ever behave in ingroup-favoring manner
without the expectation that other members of the
group behave in a similar manner? The game
theoretic analysis predicts that it is not
likely.
? Identification with the group alonei.e., the
cognition that one shares a marker with the other
members of the groupcannot make people sacrifice
their self-interest for the group.
How about the possibility that the maker works as
a surrogate of something that makes people
unconditionally cooperate?
28
Remaining Issues The marker as a surrogate of
what?
Surrogate of a kin group? Not likely The adaptive
advantage of using group-ness as a surrogate of
a kin group depends on the balance between two
types of errors. Type 1 Error To mistake a kin
member as a non-kin. Type 2 Error To mistake a
non-kin as a kin. The use of the surrogate
reduces the Type 1 error with the cost of
increased Type 2 error. The need for the
surrogate is strong when the kin members are
remotely related. The benefit of reducing Type 1
error quickly diminishes as the kin-relatedness
becomes weak. ?The advantage of reducing Type 1
error is the minimum when the need for the
surrogate is the strongest.
29
Remaining Issues The marker as a surrogate of
what?
Surrogate of a community? Not likely Unconditional
altruism for the community is fitness-reducing.
How about when the community is the unit of group
selection? Intergroup competition and the war The
incentive structure of intergroup competition, in
particular, war, is the Assurance Game rather
than a PD. It is basically a coalition formation
to win. If the coalition is likely to win,
joining is a better choice. If not, not joining
is a better choice. ?Unconditional cooperation is
fitness-reducing
30
Remaining Issues Cooperation versus Attachment?
The logical analysis tells us that unconditional
altruism toward others who share the same marker
is fitness-reducing. And yet, we seem to have
unconditional fondness toward people who share
a marker. From the logical analysis, it is not
likely that the fondness of ingroup members is a
part of the psychological mechanism that promote
unconditional cooperation (or altruism) toward
ingroup members (i.e., those who share a
marker). It may be a mechanism that make people
choose ingroup members as their interaction
partners ?attachment
31
Remaining Issues Cooperation versus Attachment?
We need to conduct a logical analysis specifying
how attachment to the group (a collection of
people who share a marker) promotes
fitness. Selective game paradigm is more
appropriate device than the forced game paradigm
to analyze the advantage of ingroup-attachment.
32
Thank you for your attention! ?