St. Lawrence University

 

 

 

 

 

Measuring Impact on Twitter

 

Journalists and Activists: Media Representation

 

 

 

 

 

 

 

 

 

 

 

 

Brett McCormack

 

Twitter & Society: SOC 4002

 

Stephen Barnard

 

5/6/16

 

 

 

 

 

 

 

 

 

 

 

Abstract

 

           When a major event rocks the world, social media explodes. Events that come to mind include the Egyptian Revolution in 2011, the Arab Spring in 2011, and more recently the death of Michael Brown in Ferguson, MO in 2014. Social media platforms like Twitter have become outlets for news and opinions for both journalists and activists to post content and information about the events following crises. After the death of Michael Brown, journalists and activists took to Twitter to share updates and news from the town of Ferguson. All parties involved are striving to get their voice heard online. The role of the media post-crisis is absolutely critical. Their actions shape what we see on Twitter. This research takes a look at what they are doing behind the scenes of posting about content that is happening online. This project also examines the tweets about media that have a significant “impact” online. The factors that make up impact in the sense that this project will be researching it include how many users the tweet reaches, how many people see a tweet, and how many people like a tweet. The greater the amount of users in the specified factor being measured means that the tweet has a bigger impact than those with less users. This research will examine these factors using the social media monitoring platform Pulsar. A question that will be addressed in this research is why/how do some journalists and activists have a bigger impact than others when it comes to tweets about media? Another question is whether there is a certain method or technique such as using just text, an image, or a link that is common among the most impactful tweets? I wanted to see what the most important elements in having a big impact on Twitter are post major event.

I hypothesized that if there are quantifiable elements/patterns among tweets that create a more of a widespread impact than others, that these points can be used to strategically have a greater impact when tweeting during future events. Bigger issues can be addressed and promoted, such as injustice and racism that were raised by the events in Ferguson. Activists can take advantage of these tips to increase their impact online. Journalists and activists can use this information to expand their viewership and reach in the long-term as well as being viewed as the go-to source for news because they now have a larger impact. “Twitter has rather evolved into a pool of constantly updating information streams consisting of links, short status updates, and eyewitness news” (Anger, 1). These new information streams come from all types of journalists and activists.

 

Who are journalists and activists?

“News organizations and journalists use Twitter as both a news source and a different media format to report stories” (Yonghwan). Journalist is defined in the Merriam-Webster dictionary as “a writer or editor for a news medium and/or a writer who aims at a mass audience.” Activist in the same dictionary is defined as “A doctrine or practice that emphasizes direct vigorous action especially in support of or opposition to one side of a controversial issue.” The work of journalists and activists is also difficult: “News can be broken on Twitter by the participants in, or observers of, a particular event. The journalist often becomes an interpreter, reacting to events quickly and frequently - and he or she often has to sift through swathes of information and opinion before deciding on what to report” (Jewell).

Journalists and activists aren’t always in agreement in full agreement with each other. There are times where activists have feelings of angst and frustration towards media. @alicesperri explained this dilemma, tweeting “Police telling media to “separate from protesters.” Protester: “let me pull out my phone, now I’m media.” This guy gets it. #Ferguson.” This tweet demonstrates activists perceptions of media, as they are not happy with the media being allowed to be on the scene while the activists and protestors have to leave. @alicesperri examined the dynamic of media treatment vs. activist/protester treatment with this tweet. This is another reason why this research wanted to look at who has the larger impact in the realm of journalist and activists interactions on Twitter.

The tweet by @alicesperri went viral, receiving 3,703 retweets and 2,676 favorites which adds up to 6,379 users total that interacted with the tweet. A possible reason the tweet may have gone viral is because the activists found commonality in the calling out of how media gets treated different than the activists. What is interesting is that according to Pulsar, the tweet came from a journalist. Perhaps the journalist was simply relaying a message or addressing a larger issue within the media-activist relationship. On Twitter, when a user retweets a tweet, all of the followers of that retweeter now see the tweet in their timeline. When a user favorites a tweet, it goes into that user's likes and all of the user's followers have the ability to see those likes. The tweet’s virality is significantly amplified and the more a tweet is retweeted and liked, it can be seen in more places on Twitter. This consumption of content can cause a cause action such a favoriting or liking which entails that a user has been influenced (Anger, 3). This influence is similar to how impact is defined in this study.   

 

What is impact?

News for the most part is about impact. TV news outlets are always trying to reach as many people as possible to catch more eyes and to establish themselves as a reliable source. On social media and more specifically Twitter however, intentions and impact are harder to measure. TV news impact can be quantified through how many people watched and at what time they watched at. On Twitter, there are many factors that shape how impactful a tweet can be. They include the amount of followers, visibility, the type of content, retweeting, reach, favoriting, hashtags, and interaction. In terms of how I will be measuring impact through Pulsar, I will be examining likes, visibility, and reach. The “likes” measure sees how many people favorited the tweet. A favorite on Twitter is a like. The “visibility” measure on Pulsar is defined as a score that considers the tweets format, where it is posted, amount of followers, and the amount of people who interact with the tweet. The “reach” measure looks at the cumulative potential audience that could have seen the content. The reach measure includes the lifespan after the tweet but does not factor in the exact times that people are online. All three measures work individually and together to determine the impact that a tweet has. This research chose to look at these three factors in order to measure how many people the tweet could have possibly reached, how many people interacted with it, and how many people liked it.  

 

Literature Review

The way that impact is measured in this study is similar to influence in literature articles. Other studies have measured influence by the amount of followers, Klout scores, and other algorithms.

The study “In the Mood for Being Influential on Twitter” by Daniela Quercia and a host of other authors looked through 31.5 million tweets and found that “...influential users tend to be individuals who express negative sentiment in part of their tweets” (Quercia). Another study done by Isabel Anger used the Klout score measure to look at influence on Twitter. Klout measure “...considers the extent to which the user’s content is “acted upon”, that is, whether it is clicked, replied, or retweeted” (Anger, 3). This is almost a mirror image of how Pulsar measures visibility. The Eytan Bakshy study “Everyone’s an Influencer: Quantifying Influence on Twitter” used an algorithm that included the number of followers, amount of tweets, and the date that the user joined to measure influence. “Reiterating that by “influence” we mean a user’s ability to seed content containing URLs that generate large cascades of reposts, we therefore begin by describing the cascades we are trying to predict” (Bakshy). All of these reports gave this research perspective and allowed me to research on the pathway to more success.

 

Methods

Using the analysis platform Pulsar, we as a class examined tweets from journalists and activists from August 9th, 2014 to November 27th, 2014. In this time period, there were 48,234 tweets. Within this set, this study examined the time period of Saturday, August 9th, 2014 to Monday, August 18th, 2014. The study chose this time period to see which users made a large impact on Twitter during the first week following the death of Michael Brown and see the users how they were able to do so. The study measured how they were able to be impactful in three ways: visibility, likes, and reach. This study also chose to filter the results within the time period by searching for the keywords “Journalist OR Activists OR Media.” I wanted to focus on tweets about media because I wanted to see how the media was represented on Twitter post major event.  

The study selected to see original posts only in order to see how original tweets made an impact. There were 198 tweets total, 142 of which were text-only, 41 containing an image, 15 with a link, and two with videos.

In order to organize the tweets by the number of likes, I sorted the tweets by tagging them in ranges. The ranges for likes were 0-50, 50-100, 100-200, 200-500, 500-700+. The ranges for visibility were 33-59, 60-86, 87-113, 114-140, and 141-166. The ranges for reach were 87-5000, 5,001-10,000, 10,001-20,000, 30,000-40,000, and 60,000-72,000+. I went through the entire 198 tweet data set and tagged each tweet for the corresponding range that it was in for likes, visibility, and reach. Each tweet now had 3 tags that explained what quantified range it fell into. I tried to make the ranges somewhat similar in size but it was challenging because there were some outliers that I will explain later on in my findings. I didn’t include a 20,001-30,000 range for reach because there were no tweets in this range. The ranges for visibility were even in size because there were no real outliers. The data for likes and reach was much more spread out compared to visibility.

This research then examined the tweets in the 200-500 and 500-700+ ranges for likes, looking at whether the tweets were text only, image based, included links, had a video, and the sentiment of the tweet. The same process was done for the 114-140 and 141-166 ranges of visibility and for the 30,000-40,000 and 60,000-72,000+ ranges of reach. The study then went on to chart these tweets by visibility, likes, and reach from highest to lowest. It is necessary to take a multifaceted approach when measuring impact of tweets on Twitter because a single variable method often excludes factors that contribute to how impactful a tweet can be. Once the results were charted, they were cross-referenced against each other by their type. The types measured were whether or not they were text only, image, link, and video.  

Findings
In determining the ranges in which to tag the tweets, it was challenging to make similar ranges for likes and reach because the tweets were so spaced out in terms of amounts and both contained outliers. The outliers for the likes ranges were above the 700 measure, and their were only two that exceeded this number. These two tweets had 2,689 and 1,393 likes respectively. I did not want to make a range for tweets with 500-2700 likes because it would not be representative of the set as a whole. I also wanted to create 5 tags of ranges per factor because more tags would be too wide of a measure and less tags seemed like it wasn’t representative enough. The tweet ranges are measured in charts below. After creating all of the ranges, I realized that I only cared about what the top two ranges for each factor (likes, visibility, and reach) contained because these tweets had the largest impact compared to their counterparts in lower ranges.

The 200-500 and 500-700+ ranges for likes had 12 total tweets. Of those 12, 7 were text only, 4 contained an image, 0 had video, and 1 had a link. 3 of these tweets had positive sentiment, 6 had negative sentiment, and 3 had neutral sentiment. The 114-140 and 141-166 ranges for visibility had 35 total tweets. Of those 36, 21 were text only, 8 contained an image, 0 had video, and 8 had links. 5 of these tweets had positive sentiment, 18 had negative sentiment, and 13 had neutral sentiment. The 30,000-40,000 and 60,000-72,000+ ranges for reach had 21 total tweets. Of those, 15 were text only, 2 contained an image, 0 had video, and 5 had links. 4 of these tweets had positive sentiment, 10 had negative sentiment, and 1 had neutral sentiment.

From these results, it was nice to see that my research aligned with the same hypothesis from the “In the Mood for Being Influential on Twitter” research, which was that there were more tweets with a negative sentiment in the top ranges than ones with positive sentiment. The text only tweets in all three categories of likes, visibility, and reach appeared more than images, links, and videos.

Limitations

There were some limitations in this study. The small sample size of 198 tweets did not represent the entire data set. Also, I don’t know how likes, visibility, and reach would work on measuring on different data set that is not about media. The sentiment measure on Pulsar is also not fully reliable, as sarcasm and lost meaning of words could calculate it falsely. I also was the only researcher with limited perspective. With more researchers, this research could have looked other factors such as follower counts to gain a fuller perspective on impact.

The course connections from class include ethnographic content analysis, Method 3, and big data in terms of quantitative vs qualitative research.

The process of tagging all the tweets was a meticulous, but rewarding process. There were times where I had forgotten to unselect tweets and accidentally tagged them in a different range. It was painstaking to go back through and find these tweets that had be wrongly tagged but satisfying when I had all 198 tweets in the set tagged correctly in their ranges for the three factors of likes, visibility, and reach.

 

Bibliography:

Anger, Isabel, and Christian Kittl. "Measuring Influence on Twitter."Proceedings of the 11th International Conference on Knowledge Management and Knowledge Technologies - I-KNOW '11 (2011): n. pag. Web.

 

Bakshy, Eytan, Jake M. Hofman, Winter A. Mason, and Duncan J. Watts. "Everyone's an Influencer: Quantifying Influence on Twitter."Proceedings of the Fourth ACM International Conference on Web Search and Data Mining - WSDM '11 (2011): n. pag. Web. 1 May 2016. <http://snap.stanford.edu/class/cs224w-readings/bakshy11influencers.pdf>.

 

Jewell, John. "How Twitter Has Helped the Emergence of a New Journalism." The Conversation. The Conversation US, 4 Nov. 2013. Web. 15 Apr. 2016. <http://theconversation.com/how-twitter-has-helped-the-emergence-of-a-new-journalism-19841

>.

Kim, Yonghwan, Youngju Kim, Joong Suk Lee, Jeyoung Oh, and Na Yeon Lee. "Tweeting the Public: Journalists' Twitter Use, Attitudes toward the Public's Tweets, and the Relationship with the Public." Taylor & Francis Online. Information, Communication & Society, 24 Oct. 2014. Web. 21 Apr. 2016. <http://www.tandfonline.com/doi/pdf/10.1080/1369118X.2014.967267>.

 

Meraz, S., and Z. Papacharissi. "Networked Gatekeeping and Networked Framing on #Egypt." The International Journal of Press/Politics 18.2 (2013): 138-66. Sage Journals. Web. 19 Apr. 2016. <http://hij.sagepub.com/content/18/2/138.full.pdf+html>.

 

Quercia, Daniele, Jonathan Ellis, Licia Capra, and Jon Crowcroft. "In the Mood for Being Influential on Twitter." 2011 IEEE Third Int'l Conference on Privacy, Security, Risk and Trust and 2011 IEEE Third Int'l Conference on Social Computing (2011): n. pag. Web. 18 Apr. 2016.