On 14 Apr 2010, at 17:09, Dan Connolly <conno...@w3.org> wrote:
Some of us have been noodling around... Open User Community Development http://www.w3.org/2008/OUCD/wiki/Main_Page I found a python implementation of the advogato trust metric, but I'm struggling to get my head around it. So I wrote a little piece of code to simulate growth of a social network; people join, and they friend/follow/certify others with certain probabilities. Also, with some probability, they joined the network to exploit it rather than to contribute; i.e. they're evil. Evil folks sometimes certify other evil folks, but contributors know better. The python code writes out each step of the simulation in JSON.
I like this approach, and expect this data-driven exploration (even if simulated data) could help get reputation-based, distributed trust out of the eternal "someday pile".
However in everyday business, sheer unadulterated evil ( Spam, fraud etc ie. generally clear-cut mischief) is one huge problem amongst many. Mailing lists, blogs etc can also suffer from over-enthusiastic participants whose contributions aren't quite right (in volume, tone or theme) for that forum. I am curious how far into that space we can go, and whether a common design can handle the spam problem and also help with fuzzier questions of authority, trust and interestingness. How far do you see it going in that direction? Is this just for 'evil people', or also 'foolish actions'?
Cheers, Dan
