How New Ideas Diffuse in Science

What conditions help new ideas spread? Can knowledge entrepreneurs’ position and develop new ideas in ways that help them take off? Most innovation research focuses on products and their reference. That focus ignores the ideas themselves and the broader ideational context. Lost are cultural conditions, concept-relations reflecting ideational use, and basic qualities of the ideas themselves. To study how new ideas spread, we use a sample of 18 million texts drawn from STEM and non-STEM fields in the Web of Science (WoS 1900-2016). We extract concepts from these texts as instantiations of ideas using a novel automated phrase mining framework (AutoPhrase) and focus on nearly 200,000 new concepts that emerged from 1992-2016. Using multi-level growth models, we predict the cumulative usage of these concepts as a function of the conditions surrounding their birth and successive usage. We find an idea’s reception is greatly affected by the evolving conditions and efforts of scholars. New scientific concepts have an identifiable “recipe” for success. They spread when they are championed by large teams of authors who reuse them; when the concept is seeded in multiple disciplines; and when the concept is dissimilar to existing terms and placed in cultural holes.