Information Technology Reference
In-Depth Information
An alternative solution can be found in weak reciprocity supported by rep-
utation. Knowing about others' past behaviors is crucial to avoid cheaters and
select good partners, especially when the group is large and it is not possible
to directly witness all the interactions. Moreover, reputation allows the costs
of social control to be reduced and distributed among individuals [12]. The im-
portance of reputation in supporting cooperation has been proven in laboratory
experiments [28], evolutionary models [20], and simulation studies [8]. Reputa-
tion is a signal that conveys socially relevant information about one's peers, and
plays a fundamental role in identyfing cheaters and isolating them.
Notwithstanding their importance in supporting cooperation, costly monetary
punishment and reputation spreading have never been directly compared. Mov-
ing from the work of Carpenter [4], in which the provision of public goods is not
negatively affected by the size of the group but by the ability of mutual mon-
itoring among agents, we enriched the original model by designing agents who
spread reputation about their previous partners. Extending previous research
on cooperation and reputation [8]and on cooperation and punishment [27], we
explore the performance of punishment and reputation as mechanisms for social
control, and we test their effects on cooperation rates both in isolation and in
combination.
Here we present a simulation platform to compare the effectiveness of costly
punishment and reputation spreading in maintaining cooperation in a popula-
tion in which defectors have a selective advantage because they exploit others'
contributions, without paying the costs of cooperation. Using a Public Goods
Game [4], we measure cooperation rates in mixed populations in which there
are pure cooperators, pure cheaters and agents who play reactive strategies. Our
contribution adds to existing literature in three ways:
1. we introduce a systematic exploration of two different social control mecha-
nisms, and we test them also in combination with other parameters, like the
group size and the costs of punishing;
2. we specify two different mechanisms for reputation:
Refuse
and
Defect
(they
will be extensively explained in Section 3).
Refuse
is a partner choice mecha-
nism which permits gossipers to avoid free-riders, whereas
Defect
is a social
control mechanism that leads gossipers to defect against non-cooperators.
Both these mechanisms are present in human societies, in which we use rep-
utational information to avoid cheaters (when this is possible), or to treat
them as we expect them to treat us. This difference between these reactions
makes it important to compare them and to understand the conditions that
make one mechanism more effective than the other.
3. we assess the extent to which reputation spreading and punishment are com-
parable mechanisms for social control, by comparing directly agents' average
contributions when costly monetary punishment and reputation spreading
are available.
In Section 2 we will introduce related work, in Section 3 we describe the simu-
lation model, and Section 4 will present the simulation results. In Section 5 we
will draw some conclusions and we will also sketch some ideas for future work.
