Reputation Systems and 4 Views

original post date:6/11/07

The Room with Four Views can be applied to reputation systems as well.

• Personal reputation (your own rank/score for somebody else)

• Peer Group reputation (what you and your friends collectively think of others)

• Empathetic reputation (asking what an individual friend thinks of third person)

• Global view (what everyone thinks of everyone, regardless of how intimately they know them)

Gradations can be applied - I know that my friend Stephen is very poor at judging the character of others, so my opinion of his opinion (the venn point between personal to peer to empathetic), despite being a close personal friend, is very low. I can almost bet that if Stephen trusts somebody, he's been conned (he's an obvious mark).

Another close personal friend knows very little about computers, so his geek reputation is low despite a high personal reputation. In my tagged buddy lists of (geeks, computer), he wouldn't even appear in the formula, but he does appear in my reputation formula for (geeks, D&D).

Yet another friend has very different tastes in a mate than I do - I'm not going to ask him if somebody is pretty, but he may know my tastes very well (empathetic view), so if he wants to 'recommend' somebody to me, I'll pay attention - with a grain of salt, because no matter how empathetic he may be, he can never truly know what turns me on (the ultimately triumph of the deeply subjective).

And I don't really give a damn what Joe Bloe thinks of the presidential races - he's distinctly average and so is his opinion. I don't really know him anyway, or any of his friends. But it may be necessary to compare my peer groups opinion to the average objective stranger 'wisdom' of crowds.

The challenge to reputation systems remains to enable this kind of complexity. Server-based (extraverted/global/collective) reputation systems ultimately fail in this scenario when compared to peer-to-peer reputation systems (triumph of the subjective) because the mathematical formula is going to be different for every user. If a reputation formula is to be meaningful to the end user, it must be easy to understand *how* a reputation was derived, and therefore it must be transparent, and customizable.

When the formula for reputation is recognized and designed as a completely subjective system, it can't be gamed - you can't game a system unless you know all of the rules of a system.

Once this complex reputation is calculated (via grid computing), it is easy illustrating the resultant reputations of everyone in my encounter list (the objective version of a buddy list). This is where the Holocene tools system comes in - adapting objective communal muspace to the personal subjective user interface by adjusting word balloon size/font size/volume/transparancy/gist level/bitrate/bittyrant/twitter frequency/avatar complexity/etc.