Researchers combined selected social media selfies and accompanying text with participants’ self-reported body satisfaction. They used artificial-intelligence systems to analyze visual and language features, then examined a smaller set of cases to help interpret the model’s predictions.

Visual features such as facial expression and body posture were more predictive among participants in the Western group, while contextual and language features contributed more strongly among participants in the East Asian group. The researchers describe these as associations in the study sample, not universal cultural patterns or evidence of cause and effect.