During the summer of 2019-20, while Australia suffered unprecedented bushfires across the country, false narratives regarding arson and limited backburning spread quickly on Twitter, particularly using the hashtag #ArsonEmergency. Misinformation and bot- and troll-like behaviour were detected and reported by social media researchers and the news soon reached mainstream media. This paper examines the communication and behaviour of two polarised online communities before and after news of the misinformation became public knowledge. Specifically, the Supporter community actively engaged with others to spread the hashtag, using a variety of news sources pushing the arson narrative, while the Opposer community engaged less, retweeted more, and focused its use of URLs to link to mainstream sources, debunking the narratives and exposing the anomalous behaviour. This influenced the content of the broader discussion. Bot analysis revealed the active accounts were predominantly human, but behavioural and content analysis suggests Supporters engaged in trolling, though both communities used aggressive language. Published at MISDOOM 2020: https://doi.org/10.1007/978-3-030-61841-4_11

1. #ArsonEmergency and
Australia’s “Black Summer”
Polarisation and misinformation in social media
Derek Weber1,2, Mehwish Nasim3,4, Lewis Mitchell4,5 and Lucia Falzon2,6
1 School of Computer Science, The University of Adelaide, Australia.
2 Defence Science and Technology, Department of Defence, Australia.
3 Data61, CSIRO, Adelaide, Australia.
4 ARC Centre of Excellence for Mathematical and Statistical Frontiers, Australia.
5 School of Mathematical Sciences, The University of Adelaide, Australia.
6 School of Psychological Sciences, The University of Melbourne, Australia.
OFFICIALMISDOOM 26-27 Oct 2020

2. Australia’s “Black Summer”
“The scale of these bushfires is unprecedented anywhere in
the world.” 1
ZDNet article2: activity on #ArsonEmergency (1-6 Jan)
OFFICIAL
1 Statement regarding Australian Bushfires, Media Release, Australian Academy of Science, 10 January 2020,
https://www.science.org.au/news-and-events/news-and-media-releases/statement-regarding-australian-bushfires
2 Stilgherrian, Twitter bots and trolls promote conspiracy theories about Australian Bushfires, ZDNet, 7 January 2020,
https://www.zdnet.com/article/twitter-botsand-trolls-promote-conspiracy-theories-about-australian-bushfires/
President of the Australian Academy of Science
Common Narrative
Hot, dry conditions & lightning,
exacerbated by climate change
Counternarrative
Arson and poor
forest management
2MISDOOM 26-27 Oct 2020

3. Research challenges
1. Identify narrative-related communities
2. Identify behavioural differences between communities
3. Study how the spread of information differs between
phases
4. Study contribution of bots to spread of information
OFFICIAL3MISDOOM 26-27 Oct 2020

4. Research challenges
1. Identify narrative-related communities
2. Identify behavioural differences between communities
3. Study how the spread of information differs between
phases
4. Study contribution of bots to spread of information
OFFICIAL4
• 31 Dec 2019 – 17 Jan 2020
• Keyword: ‘ArsonEmergency’
• 27.5k tweets by 13k accounts
Data
MISDOOM 26-27 Oct 2020

5. Polarised communities
OFFICIAL
Supporters
(497)
Opposers
(593) Unaffiliated
(11,782)
5
Retweet network
MISDOOM 26-27 Oct 2020

6. OFFICIAL6
Timeline analysis
MISDOOM 26-27 Oct 2020

7. OFFICIAL7
Community timelines
Phase 1 Phase 3Phase 2
Opposers’ peak
Unaffiliated peak
Supporters’ peak
MISDOOM 26-27 Oct 2020

8. Behavioural differences
By Community By Phase
1. Supporters active.
2. Opposers respond with RTs.
3. Supporters counter back.
Broader audience joins.
OFFICIAL8
0 5 10 15
Supporters
Opposers
Unaffiliated
0.0 0.5 1.0
Supporters
Opposers
Unaffiliated
0.0
0.5
1.0
1.5
2.0
Hashtags Mentions Quotes Replies URLs
Supporters Opposers Unaffiliated
Tweet rate Retweet rate
Other rates
MISDOOM 26-27 Oct 2020

9. Concentration of voices
(60%)
(61%)
(61%)
(61%)
(86%)
(88%)
(85%)
(83%)
(86%)
OFFICIAL9
Retweets per retweeted account
43.3 12.6 22.2
MISDOOM 26-27 Oct 2020

10. Focus of content
OFFICIAL10
Supporters Opposers
Nodes = hashtags, Links = co-mentions, Min. edge weight = 5, #ArsonEmergency removed
MISDOOM 26-27 Oct 2020

11. Spread of information
OFFICIAL
External URLs:
11MISDOOM 26-27 Oct 2020

12. Bot analysis
12
1 https://www.zdnet.com/article/twitter-bots-and-trolls-promote-conspiracy-theories-about-australian-bushfires/
2 https://botometer.osome.iu.edu/
40
#ArsonEmergency
Probably human Probably a bot
OFFICIAL
Botometer
tweetbotornot
Human Bot
MISDOOM 26-27 Oct 2020

13. Conclusion
RC1 Communities Supporters Opposers
RC2 Behaviour • More tweets
• Active throughout
• More engagement, #s & URLs
• Focused RT sources
• Broad range of topics
• Fewer tweets
• Active Ph.2 & early Ph.3
• More dissemination
• More RT sources
• Highly focused
RC3 Content
Sharing
• Mostly Narrative
• Some Conspiracy
• All Debunking
RC4 Bot
Contribution
• Highly bot-like
• Many friends/followers
• Possibly human
• Few friends/followers
OFFICIAL13MISDOOM 26-27 Oct 2020

14. Questions?
OFFICIAL14
• Derek Weber1,2, Mehwish Nasim3,4, Lewis Mitchell4,5 and Lucia Falzon2,6
• 1 School of Computer Science, The University of Adelaide, Australia.
• 2 Defence Science and Technology, Department of Defence, Australia.
• 3 Data61, CSIRO, Adelaide, Australia.
• 4 ARC Centre of Excellence for Mathematical and Statistical Frontiers, Australia.
• 5 School of Mathematical Sciences, The University of Adelaide, Australia.
• 6 School of Psychological Sciences, The University of Melbourne, Australia.
MISDOOM 26-27 Oct 2020

15. 15OFFICIALMISDOOM 26-27 Oct 2020

16. SSM
• #voteyes
• #voteno
• Used both
• Used neither
Persistent polarisation
OFFICIAL16

17. Hashtags used
17OFFICIAL

18. Hashtags used
18OFFICIAL

19. Data
• Twitter Search API
• Searches on 8th, 12th, and 17th of January
• Key term: ‘ArsonEmergency’
• Comparison searches: ‘AustraliaFire’ and ‘Brexit’
• Data
• 31st December 2019 – 17th January 2020
• 27,546 tweets
• 12,867 accounts
• 187 tweets by most active account
19OFFICIAL

20. Concentration of voices
Group Tweets Retweets Proportion RTed Accts RTs/acct
Supporters 6,972 4,224 60.5% 327 12.9
Opposers 3,587 3,184 88.8% 243 13.1
Unaffiliated 16,987 14,118 83.1% 613 23.0
(60%)
(61%)
(61%)
(61%)
(86%)
(88%)
(85%)
(83%)
(86%)
OFFICIAL20

21. Original bot analysis1
1 https://www.zdnet.com/article/twitter-bots-and-trolls-promote-conspiracy-theories-about-australian-bushfires/
30
40
#AustraliaFire#BushfireAustralia #ArsonEmergency
OFFICIAL21

22. OFFICIAL
Botometer ratings
Probably human Probably a bot
https://botometer.osome.iu.edu/
22

23. OFFICIAL
Probably human Probably a bot
https://botometer.osome.iu.edu/
23
Botometer Phase 1 ratings

24. OFFICIAL
Bot contributions
97%
<1%
2%
Tweet rates
24

25. Supporter bots
Larry
@LarryNn757
10k tweets
🚂TrumpTrain🚃 🇺🇸MAGA👊Trump2020
#Christian #Conservative #Country
#DrainTheSwamp #NOTPC
#LiberalismIsAMentalDisorder Deplorables👈
#Israel #DontTreadOnMe🇺🇸
Joined March 2018
13.2k Following 13k Followers
OFFICIAL25

26. Opposer bots
OFFICIAL26

27. Troll patterns By a Supporter on Thu Jan 09 2020 (+0000):
• 08:31:21: "@OPPOSER1 #ArsonEmergency"
• 08:31:36: "@OPPOSER1 #ArsonEmergency "
• 08:31:49: "@OPPOSER1 #ArsonEmergency "
• 08:32:04: "@OPPOSER1 #ArsonEmergency "
• 08:32:22: "@OPPOSER1 #ArsonEmergency "
• 08:32:34: "@OPPOSER1 @UNAFFILIATED1 @UNAFFILIATED2 #ArsonEmergency "
• 08:32:47: "@OPPOSER1 #ArsonEmergency "
• 08:32:57: "@OPPOSER1 @OPPOSER2 #ArsonEmergency "
• 08:33:12: "@OPPOSER1 #ArsonEmergency "
• 08:33:23: "@OPPOSER1 #ArsonEmergency "
• 08:33:51: "@OPPOSER1 #ArsonEmergency "
• 08:34:35: "@OPPOSER1 #ArsonEmergency "
• 08:35:09: "@OPPOSER1 #ArsonEmergency "
• 08:35:49: "@OPPOSER1 #ArsonEmergency "
• 08:36:16: "@OPPOSER1 @OPPOSER3 @OPPOSER4 #ArsonEmergency "
• 08:36:54: "@OPPOSER1 @UNAFFILIATED3 @UNAFFILIATED4 @UNAFFILIATED5
@UNAFFILIATED6 #ArsonEmergency "
• 08:37:10: "@OPPOSER1 @UNAFFILIATED7 @UNAFFILIATED8 #ArsonEmergency "
• 08:37:27: "@OPPOSER1 @UNAFFILIATED9 @UNAFFILIATED8 #ArsonEmergency "
• 08:38:03: "@OPPOSER1 #ArsonEmergency "
• 08:38:42: "@OPPOSER1 #ArsonEmergency "
• 08:38:55: "@OPPOSER1 #ArsonEmergency "
• 08:39:06: "@OPPOSER1 #ArsonEmergency "
• 08:39:20: "@OPPOSER1 #ArsonEmergency "
• 08:40:00: "@OPPOSER1 #ArsonEmergency "
• 08:40:12: "@OPPOSER1 #ArsonEmergency "
• 08:40:28: "@OPPOSER1 #ArsonEmergency "
27
Content
• @mention(s) #hashtag(s)
• @mention(s) #hashtag(s) URL
• #hashtag(s)
• #hashtag(s) URL
Behaviour
• Batches of similar/identical tweets
over a short period
• Repeatedly @mentioning the
same account
26 tweets at
@OPPOSER1
within 9 minutes
OFFICIAL

28. #ArsonEmergency and
Australia’s “Black Summer”
Polarisation and misinformation in social media
Derek Weber1,2, Mehwish Nasim3,4, Lewis Mitchell4,5 and Lucia Falzon2,6
1 School of Computer Science, The University of Adelaide, Australia.
2 Defence Science and Technology, Department of Defence, Australia.
3 Data61, CSIRO, Adelaide, Australia.
4 ARC Centre of Excellence for Mathematical and Statistical Frontiers, Australia.
5 School of Mathematical Sciences, The University of Adelaide, Australia.
6 School of Psychological Sciences, The University of Melbourne, Australia.
OFFICIAL