Not All Counterhate Tweets Elicit the Same Replies: A Fine-Grained Analysis

Proceedings of the The 12th Joint Conference on Lexical and Computational Semantics (*SEM 2023), pages 71–88 July 13-14, 2023 ©2023 Association for Computational Linguistics Not All Counterhate Tweets Elicit the Same Replies: A Fine-Grained Analysis Abdullah Albanyan,8 Ahmed Hassan,U and Eduardo Blanco7 8College of Computer Engineering and Sciences, Prince Sattam Bin Abdulaziz University UDepartment of Communication and Information Engineering, Zewail City University 7Department of Computer Science, University of Arizona a.albanyan@psau.edu.sa s-ahmed_hassan@zewailcity.edu.eg eduardoblanco@arizona.edu Abstract Counterhate arguments can effectively fight and limit the spread of hate speech. However, they can also exacerbate the hate, as some peo- ple may respond with aggression if they feel threatened or targeted by the counterhate. In this paper, we investigate replies to counterhate arguments beyond whether the reply agrees or disagrees with the counterhate argument. We present a corpus with 2,621 replies to counter- hate arguments countering hateful tweets, and annotate them with fine-grained characteristics. We show that (a) half of the replies (51%) to the counterhate arguments disagree with the argu- ment, and (b) this kind of reply often supports the hateful tweet (40%). We also analyze the language of counterhate arguments that elicit certain types of replies. Experimental results show that it is feasible to anticipate the kind of replies a counterhate argument will elicit. 1 Introduction Hate messages and offensive language are com- monplace in social media platforms. Twitter re- ported that more than 1.1 million accounts spread hateful content in the second half of 2020, a 77% increase with respect to the first half of the same year.1 In a recent survey of 10,093 adults in the U.S., 41% of participants reported online harass- ment on a personal level, and almost two-thirds of adults under the age of 30 reported experiencing internet harassment (Vogels, 2021). These figures, alongside other surveys,2,3 demonstrate the preva- lence of hate speech on the internet. To address this problem, the European Commission partnered with popular social media platforms to announce a "Code of conduct on countering illegal hate speech online" (European Commission, 2019), which con- tains several commitments to prevent the spread of online hate speech across Europe. 1https://time.com/6080324/twitter-hate-speech-penalties/ 2https://legalresearch.elsa.org/library/ohs/ 3https://rm.coe.int/1680700016 The enormous amount of daily data makes these platforms rely on users who manually flag hate- ful content (Crawford and Gillespie, 2016). This approach requires spending millions of dollars yearly on manual hate speech verification and mod- eration (Seetharaman, 2018). An alternative is to automatically fight hate speech by using hate speech classifiers (Section 2). However, removing users’ content—as effective as it may be—restricts free speech. According to the Pew Research Cen- ter (Duggan, 2017), “Despite this broad concern over online harassment, 45% of Americans say it is more important to let people speak their minds freely online, and 53% feel that it is more important for people to feel welcome and safe online.” A complementary strategy to address hateful content that does not interfere with free speech is to counter the hate with counterhate arguments in order to divert the discourse away from hate. Coun- …

Content truncated to 500 words