Universal facial expressions uncovered in art of the ancient Americas: A computational approach
An analysis of facial expressions in ancient Mesoamerican sculptures finds that some emotions expressed in these artworks match the emotions that modern U.S. participants would anticipate for each discernible context, including elation, sadness, pain, anger, and determination or strain. For instance, elation was predicted in the context of social touch while anger was predicted in the context of combat. The results support the hypothesis that some emotions conveyed through facial expressions are universal, reinforcing that feelings can be expressed nonverbally in ways that transcend culture. While previous studies have explored cross-cultural similarities and differences in how facial expressions convey emotions, these studies have typically asked people from Eastern or indigenous cultures to match depictions of Western expressions to situations or words in their native language. Such work may be perceived as biased since it treats Western emotional expression as the norm. To circumvent this bias, Alan Cowen and colleagues asked U.S. research subjects to label emotions expressed in ancient American art sculptures, which predated exposure to modern Western civilizations. The researchers combed through tens of thousands of images of Mesoamerican sculptures on museum websites, identifying 63 authentic sculptures that displayed facial expressions within clearly identifiable contexts, such as a smiling mother holding a baby. Next, Cowen et al. digitally separated each sculpture's expression from its context, producing, for example, one image of just the smile and one image of the mother holding the baby, with no expression visible. They asked the U.S. participants to label each image of a sculpture's facial expression with the emotion it depicted, and, separately, to label images of a sculpture's context with the emotion they would expect to see. Sculptures depicting some emotions passed the test of universality, with facial expression labels ("elated," for the mother's facial expression) matching the expectations of participants who only saw the context (an expressionless mother holding a baby). This suggests that emotional expressions can be inferred through universal human themes, such as a mother-child relationship, even without a common language. "We would eventually be interested in replicating this work in other cultures," says Cowen, noting examples of sculpture from ancient Egyptian, Indian, and Chinese cultures that could potentially be analyzed using similar study protocols. "For the time being, we are heavily focused on studying emotional expression in everyday life across many countries, aided by machine learning tools."
No comments:
Post a Comment