The Attack of the Bots and Trolls: The Social Storms that are Destroying Public Confidence in Institutions

Transcription

1 The Attack of the Bots and Trolls: The Social Storms that are Destroying Public Confidence in Institutions 19 April 2018 French

2 2017 MetricStream, Photo: James Inc. Edward All Rights Reserved. Bates/The New York Times

3 Is social technology enabling the emergence of online intermediaries, i.e., cybermediaries, that could be disruptive to the public policy process?

4 Technology and public policy Government investment in technology Government oversight of technology Government administration E government egov information systems will only achieve long term success when they sufficiently well implement interests and modes of operation of all three spheres of a governance system: formal politics, administration, and civil society. Åke Grönlund

5 Gov 2.0 e government meets Web 2.0 and social media

6 Regulation Room And the whole [of government] is cemented by giving to every citizen, personally, a part in the administration of public affairs Thomas Jefferson

7 Disrupting the public policy process Phase One

8 Value networks and disruptive innovation Performance High end Low end Source: Christensen, C. M. (1997). The innovator s dilemma: when new technologies cause great firms to fail. Boston, Mass: Harvard Business School Press. Time

9 Disruption theory at work: the case of Microsoft and the smartphone Smartphone and PC value networks (worldwide millions shipped) (Gartner, 2015; IDC, 2015)

10 Politics and Activism

11 Disruption theory in public policy Last year, more than a thousand rhinos were poached in South Africa, elephant populations have plummeted 66% in just five years, and the export of lion "trophies" has increased ten fold hunters bringing home animals' heads and bodies to stuff and mount. In response to such threats, South African Airways has declared an immediate, worldwide ban on transporting any hunting trophies made from rhinos, lions, elephants, and tigers. Now I need your support to ask Delta Air Lines to match this policy and refuse to transport exotic animal hunting trophies! Chris Green on Change.org

12 Case Study: Twitter and Scotland s independence referendum Phase Two

13 Scotland online 78% of adults had access to the internet 63% had a smartphone, and 52% had a tablet computer. Of online Scots, 66% agreed with the statement that being online is invaluable for keeping them informed about current affairs and social issues. Facebook has been used by 80% of online Scots, and 32% reported using Twitter Source: Ofcom, 2015

14 Should Scotland be an independent country? Option of more devolution of power from Westminster to Holyrood intentionally not on the ballot Sample Ballot Form (Source: UK Electoral Commission, 2014)

15 Referendum outcome: more devolution 2,001,926 (55.3%) voted No 1,617,989 (44.7%) voted Yes All three major U.K. political parties in their stumping for the No campaign promised significant additional powers for the Scottish Parliament Scotland Act 2016 additional taxing and borrowing powers for the Scottish government, and a long list of additional powers including governance over transportation and employment law, consumer advocacy, oil and gas licensing, and many other matters

16 Constraints of Twitter analysis Population that uses social media may not be proportionally representative of the general population of voters in Scotland Short document length which at the time was limited to 140 characters reduces effectiveness of sentiment analysis tools The big data technical and analytical skills of researchers black box problem Costs of Twitter access and analytics tools Lexicon based sentiment analysis does not account for images and memes Bots and troll farms

17 Research data set Full Data Set Tweet Trend (Source: Author) 1,593,013 tweets Were tweeted by someone who lives within the geographical boundaries of Scotland Contained hashtags #indyref, #bettertogether, #voteyes, or #voteno Were in a 40 consecutive day time frame, to include 3 days after the referendum

18 Hashtag analysis Twitter says Yes won VoteYes (515,430 tweets 32.4%) VoteNo (102,755 tweets 6.5%) Neutral (968,121 tweets 60.8%) NotSure (6707 tweets 0.4%) Aggregated and Filtered Hashtag Counts (Source: Author)

19 Possible bot contamination Rapid VoteYes drop after referendum VoteYes (515,430 tweets 32.4%) VoteNo (102,755 tweets 6.5%) Neutral (968,121 tweets 60.8%) NotSure (6707 tweets 0.4%)

20 Sentiment analysis of underlying issues Future of the National Health Service North Sea Oil Westminster The Pound Jobs Taxes Defense Pensions Prices Benefits

21 Sentiment polling correlation YouGov survey results in final weeks of the campaign (Source: YouGov)

22 National Health Service NHS Correlation 8/28/2014 to 9/1/2014 9/2/2014 to 9/5/2014 9/9/2014 to 9/11/2014 9/15/2014 to 9/17/2014 Average Sentiment YouGov Polling Lead for No/10

23 NHS tweets

24 North Sea oil North Sea Oil 8/28/2014 to 9/1/2014 9/2/2014 to 9/5/2014 9/9/2014 to 9/11/2014 9/15/2014 to 9/17/2014 Average Sentiment YouGov Polling Lead for No/10

25 North Sea oil tweets

26 Westminster 100% 80% 60% 40% 20% 0% 20% 40% 60% 80% 100% Westminster Correlation 8/28/2014 to 9/1/2014 9/2/2014 to 9/5/2014 9/9/2014 to 9/11/2014 9/15/2014 to 9/17/2014 Average Sentiment YouGov Polling Lead for No/10

27 Westminster tweets

28 Results Hashtag analysis not predictive of results Bot contamination Spiral of silence Some voters who sympathized with independence may have felt stronger on other underlying issues Lexicon based sentiment on several issues correlated to YouGov polling Methodology identified underlying issues most motivating to voter intent Tweet counts alone are not a predictor of correlation of issue sentiment to the YouGov polling Linear regression enabled forecasting

29 Epilogue bot usage is growing in campaigning Brexit vote (1) Two most active both fully automated Bots more active on hashtags associated with Leave Less than 1% of accounts produced 2/3 of tweets U. S. Presidential campaign (2) Political bot activity reached an all time high 18% of all presidential election related tweets Trump campaign strategically aligned bot activity to the debates By election day Trump bots were 5:1 more active than Clinton bots Sources: 1) Howard, P. N., & Kollanyi, B. (2016). Bots,# strongerin, and# brexit: Computational propaganda during the uk eu referendum (COMPROP Research Note No ). Oxford Internet Institute. 2) Howard, P. N., Kollanyi, B., & Woolley, S. C. (2016). Bots and Automation over Twitter during the US Election (COMPROP Data Memo No ). Oxford Internet Institute.

30 Epilogue fake comments on rulemaking 818,000 identical comments on FCC docket supporting rollback of Obamaera net neutrality rules Wall Street Journal

31

32 Recommendations Recommendations and Conclusions

33 Recommendations for issue and risk managers Be wary of social bots and do not use counts of social media posts as a barometer of public opinion Compare sentiment on underlying issues to polling data of voter intent to identify those issues that motivate decision making on a focal issue for voters or constituents Have a social media engagement strategy Prepare for social storms with emergency response capabilities

34 Recommendations for government officials Increase understanding of disruptive technology and effect on political and public policy processes Incorporate social technology into rulemaking more effectively Prevent bots from interfering with rulemaking I m cool with Twitter.

35 Additional Recommendations Individual change agents can use social technology to exercise their rights to petition Social media firms should enable public policy debate that is informed and reliable

36 Questions?