Machine Learning Culture: Cultural Membership Classification as an Exploratory Approach to Cross-Cultural Psychology.

Area of Science:

• Cross-cultural psychology
• Computational social science
• Cultural sociology

Background:

• Traditional research on cultural differences often relies on top-down, theory-driven methods.
• These methods can overrepresent theories like individualism-collectivism, which are based on Western observations.
• An alternative, exploratory approach is needed to capture a broader spectrum of cultural phenomena.

Purpose of the Study:

• To present and validate a machine learning-based approach for classifying cultural membership in international surveys.
• To demonstrate how interpretable machine learning can quantify cultural distances and identify differing predictors between countries.
• To offer a novel method for exploring cultural comparisons, particularly those often overlooked by traditional approaches.

Main Methods:

• Utilized machine learning models for classifying participants' cultural membership on international survey data (World Values Survey, Wave 6).
• Employed interpretable machine learning techniques, including relative variable importance and partial dependence plots.
• Constructed indices of cultural distance and analyzed specific country pairs (USA-China, USA-Japan, Japan-China).

Main Results:

• Machine learning models successfully classified cultural membership and quantified differences between countries.
• Identified key predictors that significantly differentiate cultural groups.
• Analysis revealed both established and novel dimensions of cultural difference, validating the approach's effectiveness.
• Replicated previous findings on cultural distance between the USA and China using this new methodology.

Conclusions:

• Machine learning offers a powerful, exploratory alternative to traditional methods for studying cultural differences.
• This approach effectively measures cultural distances and pinpoints specific areas of divergence between nations.
• The method holds significant potential for uncovering nuanced cultural insights in under-researched comparisons.