Brooks · iScience 2024 · cross-sectional computational behavioral study · n=5833

Deep learning reveals what facial expressions mean to people in different cultures.

Cited 17 times in the scientific literature.

Level 4 - case-series / case-control

Cross-sectional observational study with machine learning modeling (graded by design analogy for non-clinical behavioral science).

PubMed 38433918 · doi:10.1016/j.isci.2024.109175 · record verified 2026-08-26

What was done

Across six countries, 5,833 participants used a mimicry paradigm to reproduce facial expressions from 4,659 naturalistic images, generating 423,193 participant expressions. Participants rated each expression in their own language across 48 emotions and mental states. A deep neural network was trained to predict culture-specific meanings attributed to facial movements while controlling for physical appearance and context.

What was found

The model discovered 28 distinct dimensions of facial expression meaning. Twenty-one (21) of these dimensions demonstrated strong evidence of universality across cultures, while the remaining 7 dimensions showed varying degrees of cultural specificity.

Why it matters

Using large-scale naturalistic data and computational modeling, this study provides quantitative evidence that the majority of emotional dimensions conveyed by facial expressions are shared across distinct cultural groups.

Limits

The abstract does not name the six sampled countries or their cultural diversity. The mimicry task may not fully reflect spontaneous real-world facial behavior, and specific statistical performance metrics and dimension identities are not reported in the abstract.

Cited by