Information Technology Reference
In-Depth Information
5 Discussion and Future Work
In this work we proposed a direct comparison between reputation spreading and
costly monetary punishment for controlling free-riding in a Public Goods Game.
These strategies, punishment and reputation spreading, were tested in order to
investigate their effects on cooperation rates in mixed populations, and to assess
how effective they are in preventing cooperation from extinction in different set-
tings. Our contribution added to existing literature by introducing a systematic
exploration of different elements, in isolation and in combination. The effects of
these strategies in isolation have been tested in a variety of settings and environ-
ments, but it is also interesting to compare them directly. Costly punishment is
an effective strategy in promoting cooperation, and here it works quite well in
selecting out the free-riders, especially when the cost of being punished is much
higher than the cost of punishing, as expected. Even more interesting is our
finding that in larger groups the effect of punishing, even when it is not costly,
can support cooperation much better than in smaller groups.
We also introduced two different strategies based on reputation with the goal
of evaluating two alternative ways of using information about cheaters. The
Refuse
strategy can be considered pro-active and it proved to be the best strategy
in terms of cooperation rates, for every population composition. On the other
hand, the
Defect
strategy was designed as a direct retaliation against cheaters,
with negative consequences on the average cooperation rates in our populations.
These two strategies worked in different ways and are not directly comparable,
but their usage may shed new light on the conditions that make one mechanism
more effective than the other, also considering that in humans societies the two
options are usually available at the same time. It is worth noting that, on average,
cooperation rates with Refuse were close to 70%, which is a very high percentage,
especially because Gossipers were not able to affect Free-Riders' payoffs directly,
as Punishers did, so their expected e
cacy was much lower. Nonetheless, being
able to use reputation to identify Free-Riders before interacting with them was
really effecting in preventing exploitation and fostering cooperation.
This work represents a first step in a process of exploring different combina-
tions of direct, i.e., punishment, and indirect, i.e., reputation, mechanisms for
promoting cooperation in social dilemmas. Our platform will allow us to explore
single parameters and different combinations of them, with the aim of under-
standing how they work, and in what way their combination may determine the
success of a given strategy. We do not claim that reputation is a better mecha-
nism, but our data show that it is worth exploring the possibility that, for given
combinations of parameters, reputation would be more effective in sustaining
social control.
Regarding future works, with the advent of social networking platforms like
Facebook or Twitter, and motivated by the Living Labs philosophy, a possible
extension would be to perform experiments on social networks sites with real
subjects. Because of the public access to the messages shared by the users and the
underlying social network that connects then, Twitter could be the best platform
where to perform our experiment. We plan to identify active communities within
