Learning from five models beats learning from one in a controlled transmission chain

Evidence

In the 5-Model treatment, novices received images and instructions from five individuals from the previous generation. By contrast, in the 1-Model treatment, novices received the image and instructions from only one teacher from the previous generation. Keep in mind that both groups were the same size, but those in the 5-Model treatment were five times more socially interconnected . … Figure 12.1 shows the results. The average skill of each generation did not increase when people could learn from only one teacher. However, when they could learn from all five individuals in the previous generation, the average skill increased dramatically, from just above 20% to over 85%. — Joseph Henrich, The Secret of Our Success, 698

Categorization

• XREF: Mirrors the conformist-bias logic the reader tracks: sampling multiple models rather than a single authority, relevant to how many sources a coercion-vulnerable person consults. — connects to: Conformist transmission lets learners aggregate dispersed wisdom rapidly — Joseph Henrich, The Secret of Our Success: How Culture…, 179; Larger interconnected populations counteract the inherent lossiness of cultural copying — Joseph Henrich, The Secret of Our Success: How Culture…, 715; A group’s social interconnectedness outweighs individual intelligence in producing innovation — Joseph Henrich, The Secret of Our Success: How Culture…, 694 (+3 more)

Broader Connections

• Emergency call bias can transfer across agents in a multi-stage chain #expands (0.85)
• Observational learning requires attention, representation, production, and motivation #qualifies (0.75) — Albert Bandura, Moral Disengagement, 728
• Reinforcement-learning lets algorithms improve themselves faster than humans can. #complicates (0.70)