Exploring Polarization in Climate Change Debate on Twitter Aman - PDF document

<Your Name> Exploring Polarization in Climate Change Debate on Twitter Aman Tyagi Center for Computational Analysis of Social and Organizational Systems Engineering & Public Policy http://www.casos.cs.cmu.edu/ Engineering & Public Policy Motivation • Social Media is used to debate around complex social issues • Climate change is one such complex issue • Study polarizing communities debating climate change issue • Find disinformation stories in communities • Recognize topics within these communities • Bot Activity within these communities 11 June 2020 Aman Tyagi 2 1

<Your Name> Engineering & Public Policy Case Study : UN’s COP24 • United Nation’s Conference of Parties (COP) – Annual conference to discuss progress and establish obligations – COP24 – Held in Poland between 2 nd December - 15 th December 2018 • Collected ~1.3M tweets using keywords and hashtags related to Climate Change Image Source: https://www.enel.com/ 11 June 2020 Aman Tyagi 3 Engineering & Public Policy Community Detection Using Hashtags • Classify users into Believers, Deniers and Neutral • Found “#ClimateHoax” used by Deniers and “#ClimateChangeIsReal” used by Believers (Hashtags used >15 times only) 11 June 2020 Aman Tyagi 4 2

<Your Name> Engineering & Public Policy Community Detection Using Hashtags • Classify users into Believers, Deniers and Neutral • Found “#ClimateHoax” used by Deniers and “#ClimateChangeIsReal” used by Believers • Used weighted hashtag x hashtag co-occurrence network to find other hashtags • Aggregate hashtags used by each user to classify user Deniers Believers • 8,413 tweets from 2,170 Deniers climatehoax climatechangeisreal • 120,497 tweets from 15,640 Believers yellowvests ClimateActionNow qanon factsmatter 11 June 2020 Aman Tyagi 5 Engineering & Public Policy Isolating Disinformation and BOT Detection • Disinformation: – Break tweets into unigrams and bigrams – Match the unigrams and bigrams with disinformation related words/phrases • Bot Detection- CMU’s Bot Hunter 11 June 2020 Aman Tyagi 6 3

<Your Name> Engineering & Public Policy Believer and Denier Networks Denier and Believer Retweet Network 11 June 2020 Aman Tyagi 7 Engineering & Public Policy Believer and Denier Networks Denier and Believer Mentions Network 11 June 2020 Aman Tyagi 8 4

<Your Name> Engineering & Public Policy Believer and Denier Networks Denier and Believer Reply Network 11 June 2020 Aman Tyagi 9 Engineering & Public Policy Echo-chamberness of Both Groups Decreases on Adding Neutrals Echo-chamberness Echo-chamberness (Nodes) with Neutrals (Nodes) Deniers 0.007 (2,170) 0.003 (435,181) Believers 0.006 (15,640) 0.003 (432,944) • Each group is talking more to themselves 11 June 2020 Aman Tyagi 10 5

<Your Name> Engineering & Public Policy Keywords in Discussion Deniers Believer 11 June 2020 Aman Tyagi 11 Unigram Topic Analysis – Using LDA Engineering & Public Policy Deniers Believers climate climate climate cop yellowves climatechang ts cop science climate eisreal cop scam change bernie parisagre maga ement climatechang climat climate change e climateaction change globalwar sanders climate trump ming stopblahblah year like auspol climate nuclear world travel climateact carbontax ion keepitinthegr die ound ipcc climatestrike katowice industry carbon potus katowice france climateemerg fridaysforfutu climateaction talanoa ency re now world fakenews climateho month energy macron greennewdea ax tico record ambition l takeyourseat crisis people change auspol policy renewableen est ergy renewables climate leader record new planet ipcc french klimaschutz bcpoli support cdnpoli solar global know great climatech people angehoax scottmorriso sur end nmp stopadani change agw like face unfccc hoax globalwarmin cambioclim fact australia g poland 11 June 2020 Aman Tyagi 12 6

<Your Name> Engineering & Public Policy BOT Activity 100% 90% Denier Users 80% Believer Users 70% Denier Tweets Believers Tweets 60% Percentage 50% 40% 30% 20% 10% 0% 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Threshold 11 June 2020 Aman Tyagi 13 Engineering & Public Policy It’s a Conspiracy 8000 7000 Non Bot Retweeet 6000 Number of Tweets/Retweets Non Bot Tweet Bot Retweet 5000 Bot Tweet 4000 3000 2000 1000 0 Sunspot(s) Qanon/Q Illuminati Flat Earth Directed Energy Chemtrails Geo ‐ Engineering Weapons /Weather Modification 11 June 2020 Aman Tyagi 14 7

<Your Name> Engineering & Public Policy Denier’s love Conspiracies Neutral ReTweet Neutral Tweet Denier ReTweet Denier Tweet Believer ReTweet Believer Tweet 0% 5% 10% 15% 20% 25% Percentage 11 June 2020 Aman Tyagi 15 Engineering & Public Policy Example: Deniers Target Exaggerations • “entire nations could be wiped off the face of the Earth by rising sea levels if the global warming trend is not reversed by the year 2000” (Spielmann, 1989). • This story was shared 14 times, all by accounts we classified as Deniers. 11 June 2020 Aman Tyagi 16 8

<Your Name> Engineering & Public Policy Lets Call Out Each Other 6 Alarmist(s)/Alarmism 5 Denier/Hoaxer/denialist 4 Percentage 3 2 1 0 Denier Tweet Denier Retweet Believer Tweet Believer Neutral Tweet Neutral ReTweet ReTweet 11 June 2020 Aman Tyagi 17 Engineering & Public Policy Key Takeaways • For all the types of networks, we found that both deniers and believers talk within their group more than with the other group – this is truer for deniers than for believers • We found that in both deniers and believers bot- like accounts were equally active 11 June 2020 Aman Tyagi 18 9

<Your Name> Engineering & Public Policy Key Takeaways • Tweets and Retweets containing words or phrases related to conspiracies are more frequent in denier’s conversations – Conspiracies are Tweeted more than Retweeted • Exaggerations made by believers were targeted by deniers to strengthen their argument – deniers call out believers as Alarmists – believers call out deniers as Hoaxers 11 June 2020 Aman Tyagi 19 Exploring Polarization in Climate Change Debate on Twitter ORA-Demo Aman Tyagi Center for Computational Analysis of Social and Organizational Systems Engineering & Public Policy http://www.casos.cs.cmu.edu/ 10

<Your Name> Engineering & Public Policy Import Twitter Dataset 11 June 2020 Aman Tyagi 21 Engineering & Public Policy Import Dataset 11 June 2020 Aman Tyagi 22 11

<Your Name> Engineering & Public Policy Import Stance Label • Select Agent -> Editor -> Attributes -> Import Attributes 11 June 2020 Aman Tyagi 23 Engineering & Public Policy Import Attributes • Use selections as in the screenshot and click Import • In case of warning, press “Keep Results” 11 June 2020 Aman Tyagi 24 12

<Your Name> Engineering & Public Policy Delete Any Stance Category • Select Stance -> Select -> Select the category to be deleted 11 June 2020 Aman Tyagi 25 Engineering & Public Policy Delete Any Stance Category • Check “Select/Cle ar Visible” - > Select -> Nodes -> “Delete Selected Nodes” 11 June 2020 Aman Tyagi 26 13

<Your Name> Engineering & Public Policy Delete Any Stance Category • Drag and drop original meta network to make two copies • Delete Believers from one network (method similar to last slide) • Delete Deniers from the other network 11 June 2020 Aman Tyagi 27 Engineering & Public Policy Calculate Echo Chamberness • Select any one meta network and click on ”Generate Reports” • Use the setting from the following Screen Shots 11 June 2020 Aman Tyagi 28 14

<Your Name> Engineering & Public Policy • Click Finish and Check the HTML file. 11 June 2020 Aman Tyagi 29 Engineering & Public Policy Thank You! 11 June 2020 Aman Tyagi 30 15