Advancing our understanding of cultural heterogeneity with unsupervised machine learning

Abstract

National boundaries and country averages are commonly used as delimiters and proxies for culture. By doing so, not enough attention is paid to cultural heterogeneity within and overlays between countries. Deploying a Kohonen self-organizing map (SOM) as an unsupervised machine learning technique on 106,382 individual-level survey data from 66 countries, this article identifies distinct worldwide cultural prototypes, isolates dominantly occurring prototypes within countries, and uses them to calculate cultural core values. It also provides new measures for within-country cultural heterogeneity, between-country cultural differences, and cultural isolation. The results not only show the usefulness of machine learning algorithms in inductive international business research, but also have managerial relevance for international marketing and management.