Specifically, during each trial step of the QIC strategy of the experiments reported here, 100 agents are randomly partitioned into 10 teams of 10. Each agent (except the fittest agent in each team) independently selects one feature with the greatest difference in mean feature value between teammates with survival probability above and below the survival probability of the agent, and such that the consensus feature value of the fitter agents is different than the current feature value for the agent (a.k.a. local feature selection). These agents then conduct single-center trials (Np patients each) on the selected feature in their local contexts, where half of the agent's cases are tested with the selected feature value set to −1, and half tested with the feature value set to 1. QIC agents adopt the feature observed to yield higher survival, without regard to statistical significance.
http://plosone.org/article/info:doi/10.1371/journal.pone.0049901#article1.body1.sec2.sec2.sec3.p1

Can someone provide an example of what this would look like in "the real world"? Thanks.