When extremists win: On the behavior of iterated learning chains when priors are heterogeneous

How does the process of information transmission affect the cultural products that emerge from that process? This question is often studied experimentally and computationally via iterated learning, in which participants learn from previous participants in a chain. Much research in this area builds on mathematical analyses suggesting that iterated learning chains converge to people’s priors. We present three simulation studies suggesting that when the population of learners is heterogeneous, the behavior of the chain is systematically distorted by the learners with the most extreme biases. We discuss implications for the use of iterated learning as a methodological tool and for the processes that might have shaped cultural products in the real world.