The science
Twelve colleagues hold positions x_i in [0,1] on a workplace dispute. Each step they update by the Friedkin-Johnsen rule x(t+1) = alpha W x(t) + (1-alpha) x(0), where W is a row-stochastic trust matrix. With alpha = 1 (DeGroot) the team converges to the consensus pi . x(0), where pi is the stationary distribution of W: the most-trusted people carry the most influence (consensus is a weighted, not plain, average). With alpha < 1 people stay partly anchored to their initial view and a residual disagreement remains, x* = (1-alpha)(I - alpha W)^-1 x(0).
Low cross-group trust keeps the two camps apart; a mediator trusted by everyone pulls positions together. Simplifications: opinions are one number, trust is fixed, no emotion or power structure. Educational model, not conflict-resolution advice.