MunmunDeChoudhurymunmund@gatech.eduWeek13|November14,2016

CS6474SocialCompu7ng:Polariza7onandSelec7veExposure

EchoChambersOnline?:PoliticallyMotivatedSelectiveExposureamongInternetNewsUsers

Summary

•  Thepaperperformsasurveystudytoexaminehowonlinenewsreadingaffectsopinionreinforcement

•  700peoplewereaskedtoreadnewsontwonewssites–  Individualsmoreliketoreadnewsstoriesthatreinforcestheiropinions

thanthosewhichchallengethem

•  Importantfinding:–  “Havingdecidedtoviewanewsstory,evidenceofanaversiontoopinion

challengesdisappears:Thereisnoevidencethatindividualsabandonnewsstoriesthatcontaininformationwithwhichtheydisagree.”

–  Peopledon’tactivelyseektoexcludeinformationthatchallengestheiropinions,unlikewhatpopularknowledgeindicated

ExposuretoideologicallydiversenewsandopiniononFacebook

Summary

Howdidwegethere?Whatdoyouthinkwasthereasoningbehindonlineplatformspromoting/encouragingpolarizationorselectiveexposure?

Perhapsoneofthebiggeststrengthsofsocialcomputingsystemsisthattheyallowpeopletofindlikemindedothersandformcommunities.Thisseemstobeatoddswiththedangersofpolarization.

Influenceinthepoliticalsphere:62%ofadultsintheUSusesocialmediatoconsumenews,and18%ofadultsarefrequentconsumers–PewInternet

http://www.internetphenomena.com/2016/11/us-election-2016-tv-trumps-the-internet/

Figure 1: Community structure of political blogs (expanded set), shown using utilizing the GUESS visual-ization and analysis tool[2]. The colors reflect political orientation, red for conservative, and blue for liberal.Orange links go from liberal to conservative, and purple ones from conservative to liberal. The size of eachblog reflects the number of other blogs that link to it.

Because of bloggers’ ability to identify and frame break-ing news, many mainstream media sources keep a close eyeon the best known political blogs. A number of mainstreamnews sources have started to discuss and even to host blogs.In an online survey asking editors, reporters, columnists andpublishers to each list the “top 3” blogs they read, Dreznerand Farrell [4] identified a short list of dominant “A-list”blogs. Just 10 of the most popular blogs accounted for overhalf the blogs on the journalists’ lists. They also found that,besides capturing most of the attention of the mainstreammedia, the most popular political blogs also get a dispro-portionate number of links from other blogs. Shirky [12]observed the same effect for blogs in general and Hindmanet al. [7] found it to hold for political websites focusing onvarious issues.

While these previous studies focused on the inequality ofcitation links for political blogs overall, there has been com-paratively little study of subcommunities of political blogs.In the context of political websites, Hindman et al. [7] notedthat, for example, those dealing with the issue of abortion,gun control, and the death penalties, contain subcommuni-ties of opposing views. In the case of the pro-choice andpro-life web communities, an earlier study [1] found pro-lifewebsites to be more densely linked than pro-choice ones. Ina study of a sample of the blogosphere, Herring et al.[6] dis-covered densely interlinked (non-political) blog communitiesfocusing on the topics of Catholicism and homeschooling, aswell as a core network of A-list blogs, some of them political.

Recently, Butts and Cross [3] studied the response in thestructure of networks of political blogs to polling data andelection campaign events. In another political blog study,Welsch [15] gathered a single-day snapshot of the network

neighborhoods of Atrios, a popular liberal blog, and In-stapundit, a popular conservative blog. He found the In-stapundit neighborhood to include many more blogs thanthe Atrios one, and observed no overlap in the URLs citedbetween the two neighborhoods. The lack of overlap in lib-eral and conservative interests has previously been observedin purchases of political books on Amazon.com [8]. Thisbrings about the question of whether we are witnessing acyberbalkanization [11, 13] of the Internet, where the prolif-eration of specialized online news sources allows people withdifferent political leanings to be exposed only to informationin agreement with their previously held views. Yale law pro-fessor Jack Balkin provides a counter-argument7 by pointingout that such segregation is unlikely in the blogosphere be-cause bloggers systematically comment on each other, evenif only to voice disagreement.

In this paper we address both hypotheses by examining ina systematic way the linking patterns and discussion topicsof political bloggers. In doing so, we not only measure thedegree of interaction between liberal and conservative blogs,but also uncover differences in the structure of the two com-munities. Our data set includes the posts of 40 A-list blogsover the period of two months preceding the U.S. Presiden-tial Election of 2004. We also study a large network of over1,000 political blogs based on a single day snapshot that in-cludes blogrolls (the list of links to other blogs frequentlyfound in sidebars), and so presents a more static picture ofa broader blogosphere.

From both samples we find that liberal and conservativeblogs did indeed have different lists of favorite news sources,

7http://balkin.blogspot.com/2004 01 18 balkinarchive.html#107480769112109137

1 2

3

456

78

910

11

1213

1415

16

1718

19

20

21

22 2324

2526

27

2829 30

3132

3334 35 36

37 38 39

40

1 Digbys Blog2 James Walcott3 Pandagon4 blog.johnkerry.com5 Oliver Willis6 America Blog7 Crooked Timber8 Daily Kos9 American Prospect10 Eschaton11 Wonkette12 Talk Left13 Political Wire14 Talking Points Memo15 Matthew Yglesias16 Washington Monthly17 MyDD18 Juan Cole19 Left Coaster20 Bradford DeLong

21 JawaReport22 Voka Pundit23 Roger L Simon24 Tim Blair25 Andrew Sullivan26 Instapundit27 Blogs for Bush28 Little Green Footballs29 Belmont Club30 Captain’s Quarters31 Powerline32 Hugh Hewitt33 INDC Journal34 Real Clear Politics35 Winds of Change36 Allahpundit37 Michelle Malkin38 WizBang39 Dean’s World40 Volokh

(C)

(B)

(A)

Figure 3: Aggregate blog citation behavior priorto the 2004 election. Color corresponds to politi-cal orientation, size reflects the number of citationsreceived from the top 40 blogs, and line thicknessreflects the number of citations between two blogs.(A) All directed edges are shown. (B) Edges havingfewer than 5 citations in either or both directionsare removed. (C) Edges having fewer than 25 com-bined citations are removed.

excluding direct citations between the blogs (where the con-servative blogs have a clear lead), the links capture echoingof some external sources directly, but the echoing betweenthe blogs only indirectly. To measure textual similarity be-tween blogs, we identified phrases that are most informativewith respect to a background model of term frequencies inweblog data.[14]. Here we saw again a tendency, albeit aweaker one, for the same phrases to be used within commu-nities than between them.

These results suggest that although conservative bloggerstend to more actively comment on one another’s posts, thisbehavior is not accompanied by a greater uniformity in otheronline content they link to. Rather, we see both communi-ties acting as mild echo chambers by frequently discussingseparate sets of web pages and news items.

3.3 Interaction with mainstream mediaEven more common than links to other blogs are links to

news articles. Overall, the 20 left leaning bloggers cited themedia 6,762 times, while the top 20 right leaning bloggerscited media 6,364, or about once every other post on average.

0 500 1000 1500

nytimes.comwashingtonpost.com

news.yahoo.commsnbc.msn.com

nationalreview.comcnn.com

latimes.comboston.com

usatoday.comwashingtontimes.com

apnews.myway.comguardian.co.uk

foxnews.comcbsnews.com

slate.msn.comnypost.com

news.bbc.co.uktnr.com

opinionjournal.comonline.wsj.com

salon.com

# citations from weblog posts

Left

Right

Figure 4: Most linked to news sources by the top20 conservative and top 20 liberal blogs during8/29/2004 - 11/15/2004.

Figure 4 shows the most popular online news sites, andthe proportion of liberal and conservative blogs linking tothem within the top 20 liberal and the top 20 conservativeblogs. As our analysis of the home pages of the larger setof political blogs will show in Section 3.5, we find that FoxNews and the National Review receive the majority of theirlinks from the conservative weblogs, while Salon receivesover 86% of its links from liberal blogs.

Within the set of top political blogs, we also find that theNY Post, the WSJ Opinion Journal and the WashingtonTimes receive the large majority of their links from rightleaning blogs, while the LA Times, the New Republic andthe Wall Street Journal are predominantly linked to by leftleaning blogs. The remaining top-linked media sources arefairly evenly cited by the left and the right.

The actual news article citation behavior of the A-Listpolitical bloggers further differentiates the media sources at-tended to by bloggers on opposite sides of the political spec-trum. Drilling down, here are the top news articles cited byleft leaning bloggers:

1. CBS News poll of uncommitted voters shows Kerrywinning 43% to 28%

2. Sun Times article: Bob Novak predicts that GeorgeBush will retreat from Iraq if reelected

3. CBS News article on forged memos4. New York Daily News article on Osama Bin Laden

videotope, “gift” for the President5. Time Magazine poll: Bush opens double-digit lead on

post convention bounce

In contrast, the top news articles cited by right leaningbloggers are:

1. CBS News article on forged memos2. Time Magazine poll: Bush opens double-digit lead on

post convention bounce3. National Review article refuting missing explosives case4. ABC News article refuting missing explosives case5. Washington Post article on Kerry’s proposal to com-

promise with Iran on nuclear technology.

0

5

10

15

20

25

30

35

8/29

/200

4

9/5/

2004

9/12

/200

4

9/19

/200

4

9/26

/200

4

10/3

/200

4

10/1

0/20

04

10/1

7/20

04

10/2

4/20

04

10/3

1/20

04

11/7

/200

4

Date

# w

eb

log

po

sts

Right

Left

Figure 5: Time series chart: # posts discussing theCBS forged documents, right vs. left.

A time series chart further shows how quickly and stronglyconservative bloggers responded to forged CBS documents(Figure 5). The conservative bloggers saw Dan Rather’sreport as an attempt by the left to discredit President Bush.They acted quickly to debunk the report, with the charge ledby PowerLine and seconded by Wizbangblog and others. Incontrast, the pick-up among liberal bloggers occurred later,with lower volume. The most vocal left leaning bloggers onthe subject were TalkLeft and AMERICAblog.

3.4 Occurrences of names of political figuresFigure 6 shows the relative number of mentions of the

most cited political figures in our set of top 40 political we-blogs. This chart excludes George W. Bush and John Kerry,with 8071 and 8011 mentions, respectively, in order to im-prove readability. Interestingly enough, the right leaningbloggers account for 59% of mentions of John Kerry, whilethe left account for 53% of mentions of George Bush.

The chart shows that some political figures are the fo-cus of attention of primarily one side of the political spec-trum. For example, the following figures are cited by namepredominantly by the right: Dan Rather, Michael Moore,Yasser Arafat and Terry McAuliffe. On the other hand, theleft leaning bloggers account for most mentions of: DonaldRumsfeld, Colin Powell, Zell Miller and Tim Russert.

Notice the overall pattern: Democrats are the ones moreoften cited by right-leaning bloggers, while Republicans aremore often mentioned by left-leaning bloggers. (While ZellMiller is officially a Democrat, he spoke at the RepublicanConvention and has been outspokenly anti-Kerry). These

0 200 400 600 800 1000

Dick Cheney

Dan Rather

John Edwards

Bill Clinton

Michael Moore

Karl RoveMary Cheney

Al Gore

Yasser Arafat

Donald Rumsfeld

Ronald Reagan

Howard DeanColin Powell

Zell Miller

John Mccain

Tim Russert

Laura Bush

Terry Mcauliffe

# mentions

LeftRight

Figure 6: Mentions of political figures in liberalvs. conservative weblogs (excludes George W. Bushand John Kerry)

statistics indicate that our A-list political bloggers, like main-stream journalists (and like most of us) support their posi-tions by criticizing those of the political figures they dislike.An interesting topic for further study would be to comparehow balanced bloggers’ presentation of the facts are com-pared with that of mainstream media journalists.

To create this data, we ran a person name extractor overthe sets of left-leaning weblog posts and right-leaning weblogposts and counted up occurrences of the same name. Weexcluded the names of weblog authors for our top 40 weblogs.We then grouped together by hand different variants of thesame name. The person name extractor has a recall of about90%, so it is to be expected that one to three major politicalfigures are missing from the ranking.

3.5 Back to the Greater Political BlogosphereIn addition to the A-list political blogs, there are hundreds

of other blogs in the political blogosphere. We comparedtheir interests in mainstream media based on the links ex-tracted from their pages, and found these closely mirroredthose of the top 40 blogs. Fox news and the National Re-view receive 89% and 92% of their links, respectively, fromconservatives, while Salon receives 91% of its links from lib-eral blogs. The New York Times and Washington Post arethe most balanced, with the Washington Post being slightlyfavored by conservatives (55 to 40), while the NYT was al-most even, with 5 more liberals than conservatives linkingto it. We noticed that some news sources often linked to bybloggers on their sidebars are cited relatively less frequentlyin blog posts. For example, Salon receives about 60% asmany blogroll links as does the New York Times. But itonly receives 11% as many citations from blog posts, sug-gesting that its articles are not discussed as often by A-listbloggers.

Links to non-blog and non-news sites reveal other prefer-ences that the two categories of blogs have. In what follows,there are two numbers next to each site, the first being thenumber of links from liberal blogs, the second being the

Towhatextentispolarizationanewproblemwithsocialtechnologies,asagainstanofflinephenomenon?

Whatisthetentativeinteractionbetweenselectiveexposureandpeople’sbeliefsystems(wrtinformationconsumptiononline)?Canwequantifyit?

Readingopinion-reinforcingcontentcanhavewidespreadimpactonourperceptionswhatisrealandwhatisfake.Couldthisimpactourcredibilityperceptions?

IfyouweretodesignatoolthatworksintandemwithTwitterorFacebook,howwouldyouencouragediverseexposure?

Remember,humanshave“agency”,sopolarizationshouldhavesomethingtodobeyondjusthomophilyandnetworkstructure.Howwouldyouincorporateauser’sintrinsicattributestodiscouragepolarizedviewsofinformation?

http://graphics.wsj.com/blue-feed-red-feed/

Whatmakescurbingpolarizationinsocialcomputingsystemschallenging?

Adeeperquestion(fromTechCrunch):Whywould[Facebook]wanttochange?AndarepeopleevenreadyforafairFeed?