Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
121 tokens/sec
GPT-4o
9 tokens/sec
Gemini 2.5 Pro Pro
47 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Words as Trigger Points in Social Media Discussions: A Large-Scale Case Study about UK Politics on Reddit (2405.10213v3)

Published 16 May 2024 in cs.SI, cs.CL, and cs.CY

Abstract: Political debates on social media sometimes flare up. From that moment on, users engage much more with one another; their communication is also more emotional and polarised. While it has been difficult to grasp such moments with computational methods, we suggest that trigger points are a useful concept to understand and ultimately model such behaviour. Established in qualitative focus group interviews to understand political polarisation (Mau, Lux, and Westheuser 2023), trigger points represent moments when individuals feel that their understanding of what is fair, normal, or appropriate in society is questioned. In the original studies, individuals show strong and negative emotional responses when certain triggering words or topics are mentioned. Our paper finds that these trigger points also exist in online debates. We examine online deliberations on Reddit between 2020 and 2022 and collect >100 million comments from subreddits related to a set of words identified as trigger points in UK politics. Analysing the comments, we find that trigger words increase user engagement and animosity, i.e., more negativity, hate speech, and controversial comments. Introducing trigger points to computational studies of online communication, our findings are relevant to researchers interested in affective computing, online deliberation, and how citizens debate politics and society in light of affective polarisation.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (49)
  1. Mostly Harmless Econometrics: An Empiricist’s Companion. Princeton university press.
  2. Robust Hate Speech Detection in Social Media: A Cross-Dataset Empirical Evaluation. In The 7th Workshop on Online Abuse and Harms (WOAH), 231–242.
  3. SuperTweetEval: A Challenging, Unified and Heterogeneous Benchmark for Social Media NLP Research. In Bouamor, H.; Pino, J.; and Bali, K., eds., Findings of the Association for Computational Linguistics: EMNLP 2023, 12590–12607. Singapore: Association for Computational Linguistics.
  4. On the causes of Brexit. European Journal of Political Economy, 55: 301–323.
  5. Online misogyny. Journal of International Affairs, 72(2): 95–114.
  6. SemEval-2019 Task 5: Multilingual Detection of Hate Speech Against Immigrants and Women in Twitter. In Proceedings of the 13th International Workshop on Semantic Evaluation, 54–63. Minneapolis, Minnesota, USA: Association for Computational Linguistics.
  7. The pushshift reddit dataset. In Proceedings of the international AAAI conference on web and social media, volume 14, 830–839.
  8. Language (Technology) is Power: A Critical Survey of “Bias” in NLP. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, 5454–5476. Online: Association for Computational Linguistics.
  9. Enriching Word Vectors with Subword Information. Transactions of the Association for Computational Linguistics, 5: 135–146.
  10. Dogwhistles as Inferences in Interaction. In Proceedings of the Reasoning and Interaction Conference (ReInAct 2021), 40–46. Gothenburg, Sweden: Association for Computational Linguistics.
  11. Brexit and bots: characterizing the behaviour of automated accounts on Twitter during the UK election. EPJ Data Science, 11(1): 17.
  12. TweetNLP: Cutting-Edge Natural Language Processing for Social Media. In Proceedings of the The 2022 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, 38–49.
  13. Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania. The American Economic Review, 84(4): 772–793.
  14. Would your tweet invoke hate on the fly? forecasting hate intensity of reply threads on twitter. In Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining, 2732–2742.
  15. Automated hate speech detection and the problem of offensive language. In Proceedings of the international AAAI conference on web and social media, volume 11, 512–515.
  16. I Beg to Differ: A study of constructive disagreement in online conversations. In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume, 2017–2027. Online: Association for Computational Linguistics.
  17. How to disagree well: Investigating the dispute tactics used on Wikipedia. In Goldberg, Y.; Kozareva, Z.; and Zhang, Y., eds., Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, 3824–3837. Abu Dhabi, United Arab Emirates: Association for Computational Linguistics.
  18. Latent Hatred: A Benchmark for Understanding Implicit Hate Speech. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, 345–363. Online and Punta Cana, Dominican Republic: Association for Computational Linguistics.
  19. Exploring misogyny across the manosphere in reddit. In Proceedings of the 10th ACM conference on web science, 87–96.
  20. Online misinformation and vaccine hesitancy. Translational behavioral medicine, 11(12): 2194–2199.
  21. Public attitudes to the NHS. London: The Health Foundation.
  22. Analyzing the traits and anomalies of political discussions on reddit. In Proceedings of the International AAAI Conference on Web and Social Media, volume 13, 205–213.
  23. ANTi-Vax: a novel Twitter dataset for COVID-19 vaccine misinformation detection. Public health, 203: 23–30.
  24. Höller, M. 2021. The human component in social media and fake news: the performance of UK opinion leaders on Twitter during the Brexit campaign. European Journal of English Studies, 25(1): 80–95.
  25. Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692.
  26. TimeLMs: Diachronic Language Models from Twitter. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics: System Demonstrations, 251–260. Dublin, Ireland: Association for Computational Linguistics.
  27. Quantifying gender biases towards politicians on Reddit. PloS one, 17(10): e0274317.
  28. Triggerpunkte: Konsens und Konflikt in der Gegenwartsgesellschaft. (No Title).
  29. Public sentiment analysis and topic modeling regarding COVID-19 vaccines on the Reddit social media platform: A call to action for strengthening vaccine confidence. Journal of Infection and Public Health, 14(10): 1505–1512.
  30. Semeval-2018 task 1: Affect in tweets. In Proceedings of the 12th international workshop on semantic evaluation, 1–17.
  31. The ethics of COVID-19 vaccine distribution. Journal of Public Health Policy, 42(3): 514–517.
  32. An In-depth Analysis of Implicit and Subtle Hate Speech Messages. In Vlachos, A.; and Augenstein, I., eds., Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics, 1997–2013. Dubrovnik, Croatia: Association for Computational Linguistics.
  33. Social data: Biases, methodological pitfalls, and ethical boundaries. Frontiers in big data, 2: 13.
  34. Multilingual and Multi-Aspect Hate Speech Analysis. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), 4675–4684. Hong Kong, China: Association for Computational Linguistics.
  35. Comparative Evaluation of Label-Agnostic Selection Bias in Multilingual Hate Speech Datasets. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), 2532–2542. Online: Association for Computational Linguistics.
  36. The chilling: A global study of online violence against women journalists.
  37. Where should one get news updates: Twitter or Reddit. Online Social Networks and Media, 9: 17–29.
  38. Assessing the extent and types of hate speech in fringe communities: A case study of alt-right communities on 8chan, 4chan, and Reddit. Social Media+ Society, 7(4): 20563051211052906.
  39. SemEval-2017 task 4: Sentiment analysis in Twitter. arXiv preprint arXiv:1912.00741.
  40. Social IQa: Commonsense Reasoning about Social Interactions. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), 4463–4473. Hong Kong, China: Association for Computational Linguistics.
  41. Hateful Symbols or Hateful People? Predictive Features for Hate Speech Detection on Twitter. In North American Chapter of the Association for Computational Linguistics.
  42. The Economist. 2023. How a Rwandan gambit consumed the Conservative Party. https://www.economist.com/britain/2023/12/13/how-a-rwandan-gambit-consumed-the-conservative-party. Accessed: 2024-4-13.
  43. The Guardian. 2022. Decade of neglect means NHS unable to tackle care backlog, report says. https://www.theguardian.com/society/2022/dec/12/decade-of-neglect-means-nhs-unable-to-tackle-care-backlog-report-says. Accessed: 2024-04-22.
  44. The Independent. 2024. Jeremy Hunt Warned of £2bn Real-Term Cuts to NHS Funding. Accessed: 2024-04-22.
  45. Vaccine hesitancy in the era of COVID-19. Public health, 194: 245–251.
  46. Do Differences in Values Influence Disagreements in Online Discussions? In Bouamor, H.; Pino, J.; and Bali, K., eds., Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, 15986–16008. Singapore: Association for Computational Linguistics.
  47. British public opinion on Brexit: Controversies and contradictions. European Political Science, 18: 134–142.
  48. Vaccine discourse in white nationalist online communication: A mixed-methods computational approach. Social Science & Medicine, 298: 114859.
  49. Detection of Abusive Language: the Problem of Biased Datasets. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), 602–608. Minneapolis, Minnesota: Association for Computational Linguistics.

Summary

  • The paper demonstrates that trigger words significantly boost online engagement, with increases like a 13.4% rise in message frequency for terms such as NHS.
  • The paper finds that trigger words intensify animosity, spurring more negative sentiment and anger, with notable effects observed for controversial terms like Rwanda.
  • The paper analyzes over 100 million Reddit posts to offer actionable insights for moderating harmful discussions by pinpointing emotionally charged trigger words.

Understanding Trigger Words and Their Influence on Online Debates

What Are Trigger Words?

Think about the last time you scrolled through social media and stumbled upon a post that made your blood boil or inspired you to leave a long, passionate comment. Ever wondered why certain words or topics ignite such intense reactions? This paper dives right into this phenomenon with the concept of "trigger words."

Trigger words act like emotional landmines in conversations by questioning what people consider fair, normal, or appropriate. Originating from the work of Mau et al. (2023), trigger words capture moments when societal norms and individual beliefs are perceived to be under attack. In simpler terms, these words provoke strong emotional responses and drive people to react more vigorously.

Why Focus on Trigger Words?

Social media platforms, like Reddit, are flooded with debates on political and societal topics. These discussions can quickly become heated and divisive. By analyzing trigger words, researchers aim to understand how and why certain words fuel more engagement and animosity among users. In this paper, specific words tied to UK politics were scrutinized: Rwanda, Brexit, NHS, Vaccine, Feminism.

The Hypotheses

The paper operates on two main hypotheses:

  1. User Engagement: Trigger words cause an increase in user engagement, leading to a higher number of messages.
  2. Animosity: Trigger words incite more animosity, elevating levels of polarization, negativity, anger, and hate speech in discussions.

Methodology in a Nutshell

Here's a breakdown of their approach to test these hypotheses:

  1. Term Selection: They picked highly controversial terms in UK politics.
  2. Data Collection: Over 100 million Reddit posts from 2020-2022 were gathered.
  3. Control Comparisons: They compared posts across various dimensions:
    • Space: Different subreddits expected to react strongly to these words.
    • Semantics: Words with similar meanings but without the triggering effect.
    • Time: Timeframes before and after the words became trigger words.

Fascinating Findings

The paper delivered some compelling insights by analyzing levels of user engagement and animosity across these comparisons.

User Engagement: Hypothesis 1 Confirmed

Across the board, the data indicated that trigger words indeed spark more user engagement. Here's a look at some significant outcomes (% point increases in message frequency):

  • Rwanda: 10%
  • NHS: 13.4% in semantic comparisons
  • Brexit: 3.2% in semantic comparisons

It seems that certain issues - like the NHS and Rwanda - had substantial influence in driving people to engage more actively.

Animosity: Mixed but Significant Increases

While the engagement increase was somewhat expected, the findings on animosity brought to light just how polarizing these words can be:

  • Controversiality: Some trigger words increased controversiality of comments significantly (e.g., Rwanda 6.6% in semantic).
  • Negative Sentiment & Anger: Virtually all trigger words led to more negativity and anger.
    • Rwanda: Saw a 5.5% increase in negative sentiment.
    • Feminism: Anger increased by about 1.4% to 5.5% across different comparisons.

Hate Speech

Interestingly, the occurrence of hate speech showed uneven results. In areas where they used domain-specific models (like sexism for Feminism and racism for Rwanda), the results were more conclusive:

  • Rwanda: 1.8% increase in racism-related hate cases within specific subreddits.
  • Feminism: 1.4% increase in sexism-related hate comments in semantic settings.

Implications and Future Directions

This research presents deep implications for social media platforms, highlighting how specific words can drastically shift online discourse towards more contentious and damaging territory. Understanding trigger words can help platforms create better moderation policies, possibly even preemptively mitigate conflict escalation.

The Road Ahead

Regarding future developments, it's worth exploring:

  • Scalability: Extending these studies to other languages and contexts.
  • Automation: Could more advanced AI consistently identify trigger words across varying platforms?
  • Moderation Strategies: Designing AI tools to flag potential trigger words to prevent harmful escalations in real-time.

In the end, while trigger words are a small part of the vast social media conversation ecosystem, their influence is undeniably outsized. As online communication grows increasingly pivotal in our lives, so does the importance of studies like this in guiding us towards healthier discourse.

File Document Download Save Streamline Icon: https://streamlinehq.com PDF File Document Download Save Streamline Icon: https://streamlinehq.com Markdown