Keeping It All in Focus: Rethinking the Nature of Visual Framing in Protests

• In Event: New Methods for Analyzing Images-as-Data

Abstract

From news coverage to social network services, political images are all around us. In short, images matter, whether it is a briefing about COVID-19 form politicians, electoral campaigns, protests, or even wars. While it is well-know that the media deliberately chooses and highlights certain facets of events to influence the viewer’s understanding, or framing (Entman 1993; Gitlin 2003; Reese 2007), previous studies have mostly focused on textual components. Image information, however, is also important to understand how the events are framed because images tend to better capture public attention, prompt emotional and cognitive responses, and reinforce text-based messages (Brantner, Lobinger and Wetzstein 2011; Geise and Baden 2015).

Some studies have been considered visual framing (e.g., Brantner, Lobinger and Wetzstein 2011; Verse and Wicks 2006); however, they viewed images as products – images of protest or images of election campaigning. The present study is distinctive because it focuses on the process or how images can be differently filtered by common photographic choices. Building upon previous studies, this study defines visual framing as a series of choices made by journalists (and even by the public as everyday photographers) on images to influence the viewer’s attraction and responses. To examine visual framing, the current study asks the following two research questions: (1) Do people tend to choose certain photographical strategies more often when reporting on some political events as opposed to others?; And (2) are these differences predicted on the text-based contents?

While the above-mentioned research questions can be applied to various political events, this study focuses on social movement, specifically the Black Lives Matter (BLM) protests in the United States. Regarding photographic choices, this study considers angle of view (zoomed-in and zoomed-out), blurriness (blurry and nonblurry), and context (violent and non-violent). Analysis of visual framing involves (i) gathering a large set of data from the social media, (ii) applying text analysis to understand the contents, (iii) developing measurements of pictorial angles as well as blurriness, and (iv) classifying images based on the image context as well as created measurements. To achieve this goal, this study applies text analysis and various computer vision methods such as face detection, image classification, and Faster Fourier Transform.

In addition, the current study further asks (3) how the public responds to different pieces of visual information. To examine the impact of visual framing on the viewers, this study conducted online survey experiments with images manipulated in various ways. The survey included a series of questions to determine (i) the primary information respondents take from an image, (ii) their willingness to politically participate, and (iii) their evaluation of the event.

The current study makes creative and original contributions in two ways. First, this study applies computer vision methods which are proven but underutilized in political science and provide new measurement ideas. Second, it contributes to different subfields in political science. For instance, the present study addresses gap in framing effects that do not account for photographic decisions made by people, journalists, and editorial staff daily. It also speaks to political behavior and public opinion literature by proposing the impact of visual information on public attitudes and behavior. Either separately or collectively, each of contributions will further encourage interdisciplinary research (e.g., media, political communication, political behavior, and social movement) and open new research as well as substantive questions from different disciplines.

Author

• Hyein Ko, University of Illinois Urbana-Champaign