Directing the Dialogue: The Relationship Between YouTube Videos and the Comments They Spur

ABSTRACT

This study performs a content analysis of 207 YouTube videos and 45,759 comments about the Proposition 8 campaign in California. Specifically, we examine how a video's tone and focus are related to comment features. We find consistent support for the flow of information from topics mentioned in the video to topics addressed in commentary, as well as uptake of an uncivil tone from the video to the comments. Implications are discussed for promoting higher quality online information exchanges and the democratic merits of social media.

Keywords:

• Content analysis
• Internet
• politics
• YouTube
• social media

Notes

1. This dataset, as well as supplementary materials, are available at the lead author's Web site (https://sites.google.com/site/sedgerly/research/prop8).

2. These particular topics were chosen after examining the arguments presented on several Web sites supporting and opposing Proposition 8 (e.g., protectmarriage.com and NoOnProp8.com).

3. It is important to note that TubeKit retrieves up to 1,000 comments per video. This limitation, however, only impacted 12 videos out of our sample of 207 videos. Although the full comments from these 12 videos could not be analyzed, we are confident that examining 1,000 comments from a video is a robust sample size to ascertain the nature of comments about the video.

4. This variable was recoded to exclude neutral videos, leaving a total of N = 175 when video position is taken into account.

5. A complete list of dictionary words is available online at https://sites.google.com/site/sedgerly/research/prop8.

6. The measure of comment incivility does not map perfectly onto the video measure of incivility. For example, it was impractical to account for yelling in the comments, but we felt that it was an important dimension of incivility to capture in the videos. Previous research on incivility has pointed to the difficulties in transitioning the concept of incivility from the theoretical to the operational, and thus has relied on different dimensions as a proxy of the larger concept. For example, Papacharissi (2004) measured online incivility by focusing on lack of respect in comments (e.g., name-calling, hyperbole, pejorative language, profane language). Mutz and Reeves (2005) operationalized incivility in political videos through the raising of voices, interrupting, and nonverbal cues like rolling eyes and shaking heads. Ultimately, the concept of incivility is multidimensional and influenced by the medium in which the incivility occurs (e.g., text vs. video). We can be confident that the measures of video and comment incivility complement each other and fall under the common concept of incivility.

7. While the use of regression models allows for a more sophisticated analysis, another way to illustrate video uptake is to compare videos with, and without, a topic/tone and the proportion of comments that reflect that topic/tone. For example, videos that do not mention religion have about 13% of comments that do, while videos mentioning religion have about 24%—nearly double the amount. Videos that do not mention children have about 6% of comments mentioning children, while videos that do have about 12%. Videos that do not mention rights have about 16% of comments that do, while videos mentioning rights have about 20%. In terms of tone uptake, videos that are civil contain about 11% of uncivil comments, while uncivil videos contain about 16% uncivil comments.

8. We find some support for this explanation when examining media messages that were produced by the “Yes on Prop 8” campaign. For example, the leading “Yes on Prop 8” Web site connects the topics of children and rights in mentioning the need to “protect children” from “activist judges” (http://protectmarriage.com/ballot-arguments-in-favor-of-prop-8-2008).