(The) body speaks: the effects of machine-extracted body pose in image content

Abstract

The human body and its form play a vital role in nonverbal communication. However, previous literature on human images on social media has primarily studied the mere presence of humans or their facial characteristics. Likely due to its complexity and unstructured nature, only a few studies have closely examined the human body pose. This work aims to fill this gap by introducing and quantifying two body pose-related metrics—expansiveness (i.e., the space the body posture takes up) and form (e.g., standing, sitting)—using deep learning. Utilizing a large-scale fashion image database, the authors demonstrate how the proposed metrics systematically relate to one another and impact social media engagement proxied by accrued likes. They find that expansiveness and sitting (vs. naturally standing) form have significant and robust main effects on social media engagement. A subsequent online experiment establishes causality and extends the positive impact of expansiveness beyond social media engagement to consumer attitude and purchase intention toward the product presented in a given image. Dynamic imagery (i.e., perception of movement) and aesthetics are suggested as underlying drivers of the positive effects of expansiveness. Collectively, these results demonstrate how marketers can enhance social media engagement and purchase intention by manipulating a model’s pose.