When people talk more to those they agree with, do opinions converge faster?

This paper extends a classic opinion-dynamics model by letting interaction probabilities shift over time — agents grow more likely to talk to people they've previously agreed with. The key finding is that this seemingly reasonable tweak has opposite effects depending on network structure: denser networks reach consensus faster, while sparser ones take longer. The result suggests that echo-chamber-like reinforcement of existing connections isn't universally polarizing — its effect depends heavily on how connected a society already is.