Incentives - Peer-to-Peer Computing: Applications, Architecture, Protocols, and Challenges - page 125

5.4 Discussion
We have seen that in both wired and wireless systems, the major techniques
for providing incentives are: payment (virtual or real), exchange (barter), reci-
procity (pairwise), reputation (global), and game theoretic utility. Payment-
based systems work by exploiting a user's incentive in increasing or even max-
imizing its “revenue.” However, such an incentive may not be appropriate in
some practical situations. For instance, a cellular phone user may not be in-
terested in his or her income (from such sharing) but care more about the
quality of service derived from the device. Exchange-based systems fit very
well in file sharing applications because users have strong incentives for trad-
ing interested files (e.g., music files). For other applications such as forwarding
of data (e.g., in wireless ad hoc networks), it is debatable as to whether in
practice a user would be interested in exchanging data forwarding capabilities.
Reciprocity (in a pairwise manner) is similar in spirit as an exchange-based
mechanism. The major difference is that exchange is usually memoryless in
that every exchange transaction is treated as a rendezvous event, while a
general reciprocity is achieved when devices help each other during different
points in time. Thus, such a difference leads to the requirement that users
have to keep memory about the prior transactions so that users can “pay
back” each other. Due to this history-based feature, reciprocity mechanism
suffers from one drawback—a user may be out of the system when he/she
needs to pay back. Such a possible “future loss” may deter a user from gen-
uinely contributing to the community for the fear of not getting deserved pay
back. Reputation-based systems can be seen as a generalized form of reci-
procity. Specifically, while reciprocity is about a particular user pair, a repu-
tation value is a global assessment perceived by all users in the sense that the
reputation value is computed by using observations made by many different
users. By nature, similar to reciprocity mechanism, reputation systems require
substantial memory, centralized or distributed, for recording the reputation
values. Thus, it seems that such a system is more suitable for situations where
there is a persistent entity, which is logically external to the P2P system, for
keeping track of the reputation values. For example, such an entity may be a
centralized auctioneer in an electronic auction community, or the base station
(access point) in a wireless network. Game theoretic incentive mechanisms are
convincing in the sense that the utility function used can usually cover a mul-
titude of important metrics. But the problem is that it is sometimes di cult
to achieve an e cient distributed implementation of the resultant protocols.
Perhaps except in an exchange-based system which involves “stateless”
interactions, all other incentive mechanisms could suffer from tampering or
fabrication of the “incentive parameter” used: the money (virtual or real) in a
payment-based system, the reciprocity metric, and the reputation value. Thus,
additional mechanisms, usually based on cryptographic techniques, are needed