The Impact of Twitter Based Anti-Bullying Campaigns Targeted at Youth

• In Event: 3-205 - Social Media Use in Child Development Research and Outreach Efforts

Integrative Statement

As an ecological context, social media is relevant both within the macrosystem as a key feature of our current culture and within the microsystem as it is constructed within social interactions. Teens use social media as a resource to find support and information about issues in their lives (Lenhart, 2015). This study addresses three questions about Twitter anti-bullying campaigns: 1) How frequent are anti-bullying campaigns on Twitter? 2) What is the sentiment of tweets posted with an anti-bullying campaign hashtag. 3) What is the impact and duration of these anti-bullying campaigns? Given that social media has created a complex landscape for youth to connect and interact, it is important to explore how these campaigns may capture their attention and impact their perception of bullying.

From January 1, 2012 until December 31, 2014 we collected posts about bullying from the public Twitter streaming APIs (https://dev.twitter.com/docs/streaming-apis). We collected all English language posts that contained at least one of these keywords: bully, bullied, or bullying. This study focuses on a subset of those tweets that mention one of the top anti-bullying campaign hashtags and a celebrity. Machine Learning methods identified how individuals defined the celebrities in terms of their bullying role and the age of the celebrity. Combining methods from social science and computer sciences allowed for the use of massive amounts of real-world, real-time data about bullying.

Question 1. To determine the frequency of tweets related to anti-bullying campaigns first human annotators identified the top hashtags related to anti-bullying campaigns from the 500 most popular hashtags used in the data set. Twenty-eight hashtags were identified. Then machine learning techniques identified 125,387 unique posts related the campaign hashtags. On average each campaign hashtag was used publicly 9,153 times.

Question 2. To investigate whether different emotions characterized the tweets related to the campaigns, sentiment analysis was computed using Linguistic Inquiry and Word Count Software (LIWC; Pennebaker et al., 2015). The distribution of tweets by positive and negative sentiment differed significantly with a larger percentage of tweets being positive (17.3%) than negative (10.4%), however the majority of tweets were neutral (72.3%). To determine how the celebrity role in a tweet corresponded to the type of tweet we applied Machine Learning classifiers. The results show that celebrity support for an antibullying campaign is sought mainly to advocate and thereby to increase anti-bullying awareness. Additionally, a chi-squared analysis revealed that tweets where the celebrity was labeled as an advocate were more likely to be positive or neutral in sentiment than negative, X2 (1, N=125,387) = 2247.59, p=.000).

Question 3. Overall, anti-bullying campaigns on Twitter are fleeting in terms of their duration, typically only peaking for a few days, but at their peak may have a large impact among their target audience. This is especially so when a celebrity endorses the campaign, as 93.7% of the tweets classified the celebrity as an advocate. In sum, the results suggest that social media, specifically Twitter, can be leveraged to promote well-being and education about the topic of bulling among youth.

Authors

• Felice Resnik, University of Wisconsin - Madison
  Presenting Author

• Lena White, Carleton College
  Non-Presenting Author

• Amy Bellmore, University of Wisconsin - Madison
  Non-Presenting Author

• Jerry Zhu, University of Wisconsin-Madison
  Non-Presenting Author