The cumulative cultural evolution of category systems: when individual learning creates more complex artifacts

Abstract

In a laboratory experiment with human participants, we investigated the evolution of category systems in terms of their optimality and complexity over time. We contrast two evolutionary scenarios: one where category systems are passed between individuals in a serial transmission chain and another where the systems are developed within single learners. Crucially, in both cases we control for the evolutionary age of the category systems, allowing systems developed by individual learners to be iterated the same number of times as those developed among several learners (see for discussion Miton & Charbonneau, 2018). We find that both evolutionary regimes discover a similar number of unique category systems, but individual learning discovers and maintains more complex systems than cultural transmission does. Cultural transmission, on the other hand, discovers simpler systems that are optimized to be easily learnable, and therefore transmitted among different individuals with high fidelity. This result runs opposite to our field's received notion that artifact complexity is a hallmark of cumulative cultural evolution (e.g. Boyd & Richerson, 1996). There are deep mathematical similarities between evolution and learning (Suchow et al, 2017) that allow us to analyze individual learning and cultural evolution within one framework and refine our understanding of the diagnostic differences between the two.