反对外国干扰美国选举.docx

《反对外国干扰美国选举.docx》由会员分享，可在线阅读，更多相关《反对外国干扰美国选举.docx（62页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、PrefaceiiiFigures and Tables viiSummaryixAcknowledgmentsxiAbbreviations xiiiCHAPTER ONEIntroduction1CHAPTER TWOBackground and Methods 3Background 3Methods 4CHAPTER THREEFocus Groups of Partisans and Independents9How Participants Described Themselves9Russian Memes Elicited Partisan Reactions10Russian

2、 Memes Assumed to Come from Domestic Sources 11Understanding and Views of the PSA13PSA Relevance After Sourcing Revealed14PSA Might Highlight Threat of Misinformation, Election Security 17Summary of Focus Group Results18CHAPTER FOURIndividual Interviews19How Participants Described Themselves19Russia

3、n Memes Elicited Partisan Reactions19Russian Memes Assumed to Come from Domestic Sources 23Understanding and Views of the PSA25PSA Relevance After Sourcing Revealed28CHAPTER ONEIntroductionThis report outlines a strategy to counter foreign interference in U.S. elections. We posit that adversaries ar

4、e trying to exploit the fault lines that already exist within U.S. society. Although the causes of these fault lines are complex, we observe that they present tactical opportunities for our adversaries to exploit.Specifically, we observe that some adversaries seek to identify who dislikes whom and t

5、hen flood the information space to amplify these differences. The ultimate goal of these adversaries is to exploit these opportunities to reduce the probability that Americans from different backgrounds will reach consensus on issues of public concern.This is the final report in a four-part series o

6、n how to counter foreign interference in U.S. elections (Figure 1.1). The California Governors Office of Emergency Services (Cal OES) asked the RAND Corporations National Security Research Division to help analyze, forecast, and mitigate threats by foreign actors targeting local, state, and national

7、 elections.To recap, our first report outlines how Russia and other adversaries interfere in U.S. elections (Posard et al.z 2020). In it, we argue that Russian information efforts aim to elicit strong reactions and drive people to extreme positions as a way of reducing the probability that consensus

8、 will be reached. The second report focuses on who Russia appeared to be targeting on social media during the early part of the 2020 presidential election campaign (Marcellino et al., 2020). That report describes howFigure 1.1What This Series CoversDisinformation SeriesPART 1Reviews what existing re

9、search tells us about information efforts by foreign actorsPART 2Reviews what existing research tells us about information efforts by foreign actorsPART 3Identifies potential information exploits in social mediaPART 4Assesses interventions to defend against exploitsPART 5 (this report)Explores peopl

10、es views on falsehoodsL_1Russia uses fake personas (trolls) and highly networked accounts (superconnectors) to spread a variety of hyperpartisan themes effectively and quickly. In the third report, we describe how we used a survey experiment to test the impacts on the public of political content tha

11、t was sourced from Russia and its proxies (Helmus et aL, 2020). We found that Russia is successful in accomplishing a key objective from its information e任orts: eliciting strong emotional reactions and high engagement with hyperpartisan content.In our research for this final report, we invited a sub

12、set of people from our survey experiment to participate in online focus groups or one-on-one interviews. This report addresses two research questions:1. What do people think and feel about Russian-sourced content during an election year?2. Could a public service announcement (PSA) affect these views

13、?We answer both questions in four parts. First, we briefly review studies that informed our research questions, focusing on how group-based differences could create opportunities for adversaries to exploit. Second, we present results from three online focus groups made up of people who previously he

14、ld positive views of one party and negative views of the other (Democrat or Republican) or held no strong views about either party (whom we refer to as independents). Third, we present results from individual interviews with people grouped by the types of news media that they consume. Fourth, we sum

15、marize our conclusions and outline three r ec ommend at ions, based on these results, for how to mitigate foreign interference in U.S. elections.This report and the series overall support our main contention that Russia and its proxies are recycling our domestic partisanship at scale (Paul and Matth

16、ews, 2016). The content they produce and distribute, when effective, tends to elicit partisan reactions from people who might not initially define themselves in political terms. Most of the people in our sample mistakenly assumed the source of this content was fellow Americans, not Russia or its pro

17、xies. Although many of these individuals reported a positive view of a PSA warning about foreign election interference, this announcement tended to become particularly relevant to them after we explicitly told them they had just viewed partisan content sourced from Russia. We recommend that Cal OES

18、(1) monitor online activity for evidence of foreign election interference before domestic political campaigns become active, (2) release a PSA that provides a general, nonpartisan warning to the public repeatedly throughout election cycles, and (3) coordinate with social media companies to flag poli

19、tical content for attribution.CHAPTER TWOBackground and MethodsA key feature of a democracy is the capacity for diverse groups of people to find consensus on issues of public concern. In the United States, there is no shortage of these issues or dearth of divided opinions about them. Our foreign adv

20、ersaries did not create the underlying problems that created these issues, but they might, under the right conditions, exploit these tactical opportunities to further their foreign policy goals. Such exploitation is nothing new. The Soviet Union and its allies, for example, exploited race- and class

21、-based tensions in the United States during the Cold War (Ewing, 2017). One thing that is different today is that adversaries are using social media to execute their tactics.BackgroundPrevious research from this series outlined how such adversaries as Russia target elections in the United States (Po

22、sard et al., 2020), who they appear to be targeting on social media (Marcellino et al.z 2020), and the impact of these efforts on Americans (Helmus et al., 2020). We believe that reflexive control theory (RCT) is, in part, the intellectual basis for Russias recent efforts to interfere in U.S. electi

23、ons (for more details, see Posard et aL, 2020, pp. 3-4). RCT is a formal theoretical research program that first appeared in Soviet military literature during the Cold War. This program describes a way to convey information to others that could lead them to make some predetermined decision (Thomas,

24、2004). We have proposed that a key feature of Russian information efforts against the United States is to manipulate one groups view of others.We found evidence of this strategy on Twitter during the 2020 presidential election campaign (Marcellino et aL, 2020). This research revealed two types of su

25、spicious accounts whose behaviors match prior efforts by Russia and its proxies to interfere in U.S. elections: troll accounts (fake personas spreading hyperpartisan themes), and superconnector accounts (highly networked accounts that quickly spread falsehoods). Furthermore, we found that these susp

26、icious accounts targeted both conservative and liberal online communities, amplifying partisan themes and sowing divisiveness.Using known Russian-origin memes from the 2016 election, A meme, as originally defined by evolutionary biologist Richard Dawkins, 1976, referred to a behavior or style that s

27、pread from person to person. It has since been redefined, according to McCrae, 201Z as /zany fad, joke or memorable piece of content that spreads virally across the web, usually in the form of short videos or photography accompanied by a clever caption.”The two coders completed three separate tests

28、using the software package Dedoose, each with different sets of codes. The first coding test focused on self-report political affiliation, initial views of the Yosemite Sam meme, and initial views of the PSA. The second test focused on initial thoughts of the LGBT rights meme, understanding of the P

29、SA, and views of the PSA after knowing that Russia created the memes. The last coding test involved codes related to thoughts on the LGBT meme and who participants thought created the LGBT meme. We estimated intercoder reliability in Dedoose using a pooled Cohens Kappa for each test, which were 0.78

30、, 0.75, and 0.93 respectively (de Vries et al., 2008). our third report conducted a randomized controlled trial to test the impacts of this content on a sample of Americans (Helmus et al.z 2020). This third report describes evidence we found that Russian content elicits emotional reactions from user

31、s that correlate with their selfreported willingness to publicly like and share the content. In addition, telling users that Russia created the content attenuated these reactions.Building on this prior work, we conducted a series of focus groups and interviews of volunteers from the survey experimen

32、t to hear, in their own words, their thoughts and feelings about select Russian memes. We then asked them for input on a general warning about broader information efforts by foreign actors. In the next section, we describe the methods for these discussions.MethodsThe focus groups were completed onli

33、ne via Facebook; we saved the chat transcripts. The individual interviews were telephone calls that we recorded; we then used Amazon Transcribe to auto-transcribe the audio. Two members of the research team checked the accuracy of each auto-transcribed transcript against the audio recording.Next, we

34、 constructed a coding guide for these transcripts that we based on our interview protocol (see Appendix A). We first conducted a trial for coding in which two team members used Dedoose to independently code each of the focus group transcripts. To ensure intercoder consistency, these two team members

35、 discussed their coding decisions and streamlined the codes in the coding guide, then conducted three separate coding tests in Dedoose for relevant codes from the individual interviews.We selected the Yosemite Sam meme and LGBT rights meme because both elicited the largest partisan response by respo

36、ndents in our survey experiment.Participants then answered questions about the memes. Figure 2.1 displays three images seen by participants: a conservative Yosemite Sam meme (sourced from Russia), a liberal lesbian, gay, bisexual, and transgender (LGBT) rights meme (sourced from Russia), and a PSA?F

37、igure 2.1Russian Memes and Public Service Announcement Used in This StudyI WAS BANNED FROM TELEVISION FDR BEING TOO VIOLENTLIKE B SHARE, IF YDU GREW UP WATCHING ME -ON TELEVISION, HAVE A GUN, AND HAVENT/ 7 SHOT DR KILLED ANYONE!、During the 2020 election season, foreign actors and cyber criminals are

38、 spreading false and inconsistent information through various online platforms in an attempt to manipulate public opinion, discredit the electoral process, and undermine confidence in U.S. democratic institutions.RECOMMENDATIONSSeek out information from trustworthy sources, verify who produced the c

39、ontent, and consider their intent.Rely on state and local election officials for information about voter registration databases and voting systems.View early, unverified claims with a healthy dose of skepticism.Verify through multiple reliable sources any reports about compromises of voter informati

40、on or voting systems, and consider searching for other reliable sources before sharing such information via social media or other avenues.Report potential election crimessuch as disinft)rmation about the manner, time, or place of voting to the FBI.If appropriate, make use of iii-platibnn tools offer

41、ed by social media companies for reporting suspicious posts that appear to be spreading false or inconsistent information about voter information or voting systems.The sample of participants for our focus groups and individual interviewees are people who completed our survey experiment on Russian di

42、sinformation (for more details, see Helmus et al.z 2020). Of 1,529 people who completed the second wave of this survey experiment, 897 volunteered to have a follow-up contact for our focus groups or individual interviews (58.7 percent). Of these volunteers, we identified 106 people who met our inclu

43、sion criteria for the focus groups and another 441 people who met our criteria for inclusion to be part of our interview sample. We then used quota sampling for both samples and stopped sending invitations after we met our quotas for the focus groups (n = 15) and individual interviews (n = 27).Sampl

44、e of Focus Group ParticipantsCandidates for this sample completed two questions on this survey that asked them how favorably or unfavorably they viewed Democrats and Republicans. Using these answers as our criteria, we constructed three separate focus groups that were held on Facebook in October pri

45、or to the 2020 presidential election. Here, we provide descriptions for each of our focus groups:1 Republican Group: This focus group had eight participants. Five reported holding a very favorable view of Republicans and a very unfavorable view of Democrats. Three reported having just a favorable vi

46、ew of Republicans and a very unfavorable view of Democrats. There were not enough people from our survey experiment who volunteered for us to follow up with them and who self-reported both very unfavorable views of Democrats on our survey experiment and very favorable views of Republicans. Thus, we

47、adjusted this criteria to include participants self-reporting they held just favorable views of Republicans and a very unfavorable view of Democrats. This decision likely reduced the degree of partisanship within the Republican focus group.2 Democratic Group: This group consisted of ten participants

48、 who reported holding a very favorable view of Democrats and a very unfavorable view of Republicans.3 . Independent Group: This group had a total of nine participants who reported having neither a favorable nor unfavorable view of Republicans or Democrats.Sample of Individual IntervieweesNo intervie

49、wees in this sample participated in our focus groups. In the survey experiment from which these interviewees were drawn, participants were sorted into five distinct groups based on their news consumption types. For the interviews, we selected participants from the following three groups: partisan left, partisan right, and analytical skeptics. The partisan left and right groups were defined by their partisan beliefs; each group represe