Proceedings of the Annual Meeting of the Cognitive Science Society

Social learning is often valued for reducing the costs of individual exploration and avoiding mistakes. Its benefits, however, extend beyond simple error prevention in domains where knowledge is compositional: when ideas are generated by combining existing elements, meetings of minds are a fertile ground for novel innovations. We introduce an evolutionary agent-based model in which agents pursue diverse learning strategies. Some agents create knowledge independently, others exchange and recombine ideas. Agents interact repeatedly in a compositional task environment, building, sharing, and combining knowledge components. We find that social interaction produces a rich pool of partial ideas, and recombination among these ideas can connect disconnected knowledge strings, though only alongside accurate discoveries made by individual learners. Social recombination allows populations to explore combinatorial solution spaces more effectively than individual learning alone, or success-biased social learning. By highlighting the interplay between idea exchange, recombination, and selective copying, our results reveal a novel pathway through which social learning enhances adaptive knowledge accumulation.