Humans worldwide make similar facial expressions in various social situations, shows a Nature paper. An analysis of more than 6 million videos suggests substantial universality in how people across the world express themselves in various social situations. However, the findings do not determine whether emotions themselves are universal; more research is needed to get a fully unbiased view of how emotions are expressed in everyday life.
Do people across the world smile more during weddings and frown more during funerals? Whether such expressions are preserved across cultures has been challenging to assess, and previous research using survey-based approaches—constrained by language barriers and small samples—has led to diverging viewpoints regarding the universality of emotion. In this study, Alan Cowen and colleagues used a deep neural network (DNN), a type of machine learning, to assess real-world behaviour and ascertain whether social contexts are associated with specific facial expressions across different cultures.
English-speaking people in India trained the DNN to identify 16 patterns of facial movement associated with distinct English-language emotion categories. The DNN then learned to cluster similar patterns as individual expressions (such as a smile) and assessed 6 million YouTube videos from 144 countries, and found that similar expressions often occurred in similar contexts worldwide. For example, facial expressions often labelled as ‘awe’, ‘contentment’ and ‘triumph’ were associated mostly with weddings and sporting events across different regions. Additionally, each type of facial expression had distinct associations with a set of contexts that were 70% preserved across 12 global regions, suggesting considerable universality across the world.
These findings have important implications for understanding the origins, functions and universality of emotion. In an accompanying News & Views Article, Lisa Feldman Barrett notes that the present study identifies associations between expressions and social situations in more-natural settings than previous studies. She suggests that future work should use more diverse cultural groups to train the DNN, to avoid depending solely on English speakers’ beliefs and stereotypes about emotional expressions.
Criminology: Predicting police enforcement bias in major US citiesNature Human Behaviour
COVID-19: Assessing instances of long COVID in UK health dataNature Communications