AUTHOR
Stephanos Androutsellis-Theotokis
ABSTRACT
Distributed computer architectures labelled “peer-to-peer” are designed for the sharing of computer resources (content, storage, CPU cycles) by direct exchange, rather than requiring the intermediation or support of a centralized server or authority. Content distribution is the dominant peer-to-peer application on the Internet, allowing personal computers to function in a coordinated manner as a distributed storage medium by contributing, searching and exchanging digital content.
A decentralized peer-to-peer content distribution network can be thought of as a community of users, not unlike any other social group of people with common interests or functions. In our case, each member of this community has two goals: A “personal” goal, which is to benefit the most from the services the community can provide for them (in our case by obtaining the most desired digital content at the lowest “cost”), and a “global” goal, which is to help the entire community to continue existing, prosper and proliferate, in part through the services they can provide to the other members (in our case the content, storage and bandwidth they own or control.
These two goals are often conflicting and, in the absence of any centralized authority, control or enforced rules, they govern the social behaviour of each member and ultimately of the entire community. The general rule (and paradox) here is that universal cooperation by all members is best for the overall good of the community, however if no specific incentives are enforced, the users who directly benefit the most are those who take advantage of the content provided by others without contributing anything themselves. The “prisoner’s dilemma” metaphor can be used to study the extent to which the owners of intellectual property are willing to share with others.
In the on-line content distribution community described above, one is afforded the opportunity to actively interfere by introducing and applying a variety of incentive mechanisms, to enforce a more collective behaviour and eliminate the “freerider” effect. We introduce and discuss the categories of incentive mechanisms that can be (and have been) used, the technologies on which they are based and ultimately the cost associated with such measures.
We proceed to focus our discussion on the issue of reputation mechanisms, which attempt to be technological solutions to the problem of controlling social behaviour in peer-to-peer content distribution networks, by ensuring that users with poor social behaviour will be less privileged and will have fewer rights than users who work more towards the common good. The two main problems these mechanisms are faced with in their application to peer-to-peer architectures is the lack of fixed identification for the users and the lack of a history of repeated interactions between users due to the typically large sizes of the networks. We discuss various reputation mechanisms, their related advantages and costs, and the way in which incentive mechanisms can be based on the private or shared history of users which is assembled by the reputation mechanisms. We finally examine in more detail the particular issue of anonymity, which presents an interesting “inverse incentive” mechanism for single users as, in order for their anonymity to be protected, they need to ensure that large crowds are maintained and operate within the content distribution network. They are therefore inherently motivated to work for the “global good”, since it will directly reflect on themselves by offering improved anonymity.
We conclude our arguments by reversing our metaphor, briefly comparing the “virtual” file sharing community of our discussions to a “real” social group of people, and discussing how corresponding incentive and reputation mechanisms work in the second.