Add or Change? Photos Make Smartphonegenerated Reviews Different: Evidence from an Online Platform

Expressing opinions through photographs shared online, becoming a prominent form of content creation, so it is crucial to expand our understanding of how photos affect the popularity of user-generated content on social media. In this work, by applying cognitive cost theory and computer vision method to UGC, the mediation role of photos between device (smartphone and PC) and perceptual content as well as information richness in online reviews is investigated. We applied LIWC automated sentiment-analysis tool to mine perceptual content from review text and used histogram of oriented gradient (HOG) and support vector machines (SVM) to mine additional information from photos, verifying the interaction between photos and text. The empirical findings show that compared with PCs, smartphones tend to have more photos attached to reviews, which, in turn, enhances the perceptual content and information richness of the reviews, enabling reviewers to enrich and modify the conceptual gist of their experiences.