Market Reputation Systems-
These are sensible suggestions and are an element of ebay's system - however this is just a start - and it is clear that there is a conflict and trade off between Ease and understandability and a system being difficult to game - but this might not be fatal, as I outline below.
At the recent MIT online reputation symposium we discussed ease and publicness with the ebayers. They spend a lot of effort dealing with buyers and sellers concerned about their reputations and their feeling was that it was critical to make it easy and easy to understand, "to figure in one's head" - which restricts the ability to do second and third level reputation calculations but also rules out some of the more complex non-gamability systems.
I think this notion of Public should be taken further to a system of reputation stores and reputation APIs, just as the nimble blogging community is doing with systems such as Technorati. One difficulty of existing reputation systems is that they are under the direct control of the marketplace that produces the data - making tracking reputations a costly, and non-core, part of their businesses. By releasing ones reputation data through an API you relieve yourself of this burden and create a market for innovation as people cut-and-slice the data in different and interesting ways. We should be thinking of an xml schema and unifed API for reputation systems.
The third point, "Difficult to Game", was one that came up often at the MIT symposium where the Computer Scientists usefully termed this, "incentive compatibility", ie the system needs to be compatible with people's incentives. Quite complex systems are quickly established and many formal methods brains working on this non-gamability problem - most of which are clearly in conflict with both Ease and probably Publicness, yet I think that this need and the conflicts it creates are overblown.
By analogy to the non-wired market one's reputations are important but far from non-gameable and the system still works fine. In fact when thought of in the same terms that one uses for virtual markets it is a damn risky proposition doing business in the real world.
So what makes it work? There are many institutions that underpin the economy and a critical one is insurance. Insurance is critical to the business we do everyday - who would drive if one didn't think that others had insurance? Similarly insurance underpins the transport trade systems and real-estate investments that we make. We use options and derivatives to provide insurance to our capital investments and income-interruption insurance for the labour market. In this way we may not totally trust the people that we do business with but we do trust the system - we are assured that the system, with the checks and balances of insurance, is safe enough for business.
So to the system designers for online reputation systems designed to ease the transition to digital markets I say it is possible to make your systems totally non-gameable, but at what computational costs and at what level of complexity? Insurance, along with the other institutions that underlie the market, will be part of a hybrid solution that builds, not perfect trust, but system assurance.
Posted by james at May 21, 2003 06:44 PM