Found an interesting paper presented at HICSS (Rodriguez, 2007) talking about a social network system for collective decision making. Basically (to avoid reading the paper), the authors developed a social network in which users could express different degrees of trust for each other. The system could be used to make a collective decision on a posed question, e.g., “What should be done in xxx situation?”
They tested three algorithms, each of which was aimed for a different dynamic:
1) Direct democracy (everyone gets a vote, if you don’t vote your vote is lost)
2) Dynamic distributed democracy (everyone gets a vote, if you don’t vote it passes to a person you trust; if they don’t vote it passes to a person they trust; onwards until it reaches someone who votes whose vote is then worth two)
3) Proxy (expert) network (everyone gets votes proportional to their in-degree trust links, otherwise same as 2)
The actual algorithm was based on particle swarms to make it more probabilistic and graded, but basically the same as described above. Turns out that in the test problems all of the three forms led to very similar answers, but that might be an issue with the problems not exposing the differences.
It is an interesting paper with some potential design patterns for collaborative intelligence.