Quantifying Controversy in Social Media

Abstract

Which topics spark the most heated debates in social media? Identifying these topics is a first step towards creating systems which pierce echo chambers. In this paper, we perform a systematic methodological study of controversy detection using social media network structure and content. Unlike previous work, rather than identifying controversy in a single hand-picked topic and use domain-specific knowledge, we focus on comparing topics in any domain. Our approach to quantifying controversy is a graph-based threestage pipeline, which involves (i) building a conversation graph about a topic, which represents alignment of opinion among users; (ii) partitioning the conversation graph to identify potential sides of the controversy; and (iii) measuring the amount of controversy from characteristics of the graph. We perform an extensive comparison of controversy measures, as well as graph building approaches and data sources. We use both controversial and non-controversial topics on Twitter, as well as other external datasets. We find that our new random-walk-based measure outperforms existing ones in capturing the intuitive notion of controversy, and show that content features are vastly less helpful in this task.

Exploring Controversy in Twitter

“…This hypothesis is based on the fact that a controversial topic entails different sides with opposing points of view, as well as on previous evidence that individuals on the same side tend to endorse and amplify each other's arguments [1,2,3]. We studied this hypothesis in previous work [4], and found strong evidence that the retweet graph of a controversial topic presents a clustered structure that reveals the opposing sides of the debate. Moreover, in the same work, we developed a random-walk-based measure that quantifies accurately how controversial a topic is by taking into account the structure of its retweet graph.…”

“…The main limitation of previous work is that the majority of studies have focused on known, long-lasting debates, such as elections [1,3]. Our system is the first one to attempt controversy detection in the wild, on any topic, and without human data curation [4].…”

“…The final output of the pipeline is a value between zero and one that measures how controversial a topic is, with higher values corresponding to higher degree of controversy. We provide a high-level description of each stage here, for further details refer to the original work [4].…”

Telling Apart Tweets Associated with Controversial versus Non-Controversial Topics

“…This service has also been widely used to disseminate information during emergencies and natural disasters, and to mobilize support for social and political movements (Lotan, Graeff, Ananny, Gaffney, & Pearce, 2011). As with many other outlets of public opinion, Twitter features the emergence of polarization around controversial issues (Addawood & Bashir, 2016;Garimella, De Francisci Morales, Gionis, & Mathioudakis, 2016), and provides a forum where people can express their opinions, which may be conflicting (Pennacchiotti & Popescu, 2010). This paper focuses on the classification of tweets on topics that are perceived as controversial versus non-controversial.…”

“…Previous research used Twitter for detecting both controversy and controversiality (Conover et al, 2011;Garimella et al, 2016;Pennacchiotti & Popescu, 2010). To date, much of the previous research on controversiality has used data from political debates (Adamic & Glance, 2005;Conover et al, 2011;Mejova, Zhang, Diakopoulos, & Castillo, 2014;Morales, Borondo, Losada, & Benito, 2015), news (Awadallah, Ramanath, & Weikum, 2012;Choi, Jung, & Myaeng, 2010;Mejova et al, 2014), and social media, such as blogs (Adamic & Glance, 2005), and Wikipedia (Dori-Hacohen & Allan, 2013;Kittur, Suh, Pendleton, & Chi, 2007;Rad & Barbosa, 2012).…”