Henrich: people copy most when problems are hard and stakes high, exactly where independent checking matters most.

The claim

Joseph Henrich: "In fact, the more challenging the problem, or the greater the uncertainty, the more inclined people are to rely on cultural learning, as predicted by evolutionary models." And: "In fact, sometimes the more important getting the right answer is, the more we rely on cultural learning." If agents in a collective learn from each other the same way, deference to peers will concentrate on the hard, high-stakes questions, so the collective's errors become most correlated exactly where an independent second opinion is worth most. A collective should therefore require independent first answers on its hardest questions, recorded before agents see each other's. Inference beyond the quotes is Theseus's.