Simulating the evolution of consensus in Wikipedia, for some reason

An Agent-Based Model of Wikipedia Edit Wars : How and When Consensus is Reached

A very strange but interesting paper which tries to model how quickly consensus is established by editors of controversial Wikipedia articles. I can't help feeling that there's an awful lot of complex modelling gone in to producing conclusions which for the most part are obvious and common-sense. At the same time, trying to model consensus-formation is very interesting, even if this particular case is oddly specific. It does have the option of calibrating against well-documented examples though.

The model includes a bunch of different parameters :
- The ability of agents to commit changes or reversions to articles, which can be both positive or negative in favour of a given viewpoint (hence four possible actions they can take)
- The credibility of different agents
- A probability that agents will actually take one of the possible actions
- A "payoff" given to agents for taking actions, which varies in a highly complex way based on the agent's stance and the other agent's actions and credibility
- A desired level of payoff for each agent
- The ability for agents to "learn", in the sense that the probability of the actions they take vary based on their results, though I can't make head nor tail as to how this is actually implemented

All of this results in the rather uninteresting conclusions :
- The more likely an agent is to make a revision, the longer it takes to establish consensus
- Consensus is established more rapidly when credibility of the agents involved is higher, up to a point
- Consensus is established most rapidly when everyone agrees and most slowly when opposing sides are about equal (well, blow me down with a feather (!))

It's still an interesting idea. I'd like to see it applied to more general situations.

http://casos.cs.cmu.edu/publications/papers/2015AnAgentBasedModelofEditWars.pdf