Exploring Organizational Self-(re)presentations on Visual Social Media: Computational Analysis of Startups’ Instagram Photos Based on Unsupervised Learning

Abstract

Presenting oneself visually is an essential part of online behaviors on social networking sites (SNSs), and this is also the case for organizational accounts. Examining organizational self-presentation on their SNS accounts can show the key instruments through which organizations communicate online with the public. Thus, taking startups’ Instagram accounts as examples, this study aims to explore what and how organizations present visually on the photos uploaded to their Instagram accounts. The photos were analyzed in terms of their content, texts in photos, and facial and pixel-level characteristics. Photos and accounts were clustered respectively and resulting clusters were compared as well. The results suggest that text and people are the major objects represented in startups’ Instagram photos. The text presented in the photos was mainly about startups’ efforts to realize social values that can benefit communities by aiding people to do something in a timely manner. Two photo clusters were found: one mainly represented people and people-related words and the other mainly presented texts. Two account clusters were also found: one mainly consists of IT companies that presented texts in their photos and the other mainly consists of human-related services that presented people in their photos. This study can contribute to the body of literature by identifying the visual characteristics of organizational SNS photos and expanding self-presentation research to organizational SNS accounts.