You are on page 1of 19

RachelCohen

CanTwitterAccuratelyPredictElections?
Withinthepastfewyears,associalmediahasbecomeincreasinglypopular,thepredictivepower
oftwitterhasbecomeaverytalkedabouttopicwithinthePoliticalSciencesandCampaign
Managementfield.Accordingtosome,correctlyandcarefullyanalysingtwittercanleadto
researchersbeingabletopredictanythingfromboxofficesalestodifferentepidemicshowever,
otherswillsaythattwittercannotpredictanythingbetterthansimplechanceisabletopredict.

Thispaperisabouttwitteranditsabilitytoaccurately,orinaccurately,predictelections.Thisis
notaboutanyotherformofsocialmediaoranyothertopicthatitmightbeabletopredict
becausethatisbeyondthescopeofthisconversation.Thoughsomeofthearticlescoveredspan
differentcountries,themajorityofthemrelatedirectlytopresidentialelectionswithintheUnited
States.

Therearemanyflawswhenitcomestocurrentresearchconversationsaboutwhetherornot
twittercanbeusedasapredictivemethod,mainlybynotaccountingforvariousfactorsthat
mightaccountforswayedorbiasedinformation.Forfutureresearchtopossiblybecomemore
effective,morecontrolledvariablesmayhavetobestudiedtoensurethatnooutsideinformation
couldmessupresults.

How(Not)ToPredictElectionsMetaxas,P.T.,Mustafaraj,E.,andGayoAvello,D.2011.In
ProceedingsofPASSAT/SocialCom.

Therearethreeauthorsofthisarticle.ThefirstisProfessorPanagiotisTakisMetaxas,a
professorofcomputersciencescurrentlyworkingmoreintothefieldofsocialmediaand
andcrowdsurfingonsocialplatforms.ThesecondisaProfessorattheUniversityof
Oviedo,whoworksinthecomputersciencesdepartmentbuthasrecentlybeendoing
studiesbasedaroundsocialmedia.ThethirdisProfessorEniMustafaraj,alsoaprofessor
inthecomputersciencesfield.Shehighlightsonherwebsitethatasciencemagazine
publishedProfessorMetaxasandherarticleentitledSocialMediaandtheElectionsin
2012.Thefirstandthirdauthorshavepublishedatleastonemorearticletogetherwhich
ismentionedbelow.

Thisisoneofthefewpapersthatcastsdoubtonthewidespreadopinionthatsocialmedia
canbeagoodtoolusedtopredictdifferentelectionsandotherthingswithinpolitics.
Afteranalysingmanydifferentstudies,theauthorsconcludedthatanalyzingtwitterand
thethingsthatpeoplepostontheregivesonlyaslightlygreaterlikelihoodofpeopleable
topredictapoliticalelectionthanchancedoes.Thisessentiallymeansthateverythingcan
bepredictedtosomemeasurebasedoncommonsenseorpublicopinionbynewssites
andusingtwittertopredictthisonlyhelpsbysuchalittlebitthatitispractically
negligible.

Thisarticlealsoaddressesthreemajorpointswhenitcomestopeoplewhoclaimthat
twittercanbeusedsuccessfullyasapredictivemeasure.Thefirstpointwasthatthere
needstobeaclearlydefinedlogarithmwhich,accordingtothis,thererarelyis.The

secondwasthatitneedstotakeintoaccountthatthepartofthevotingpopulationthat
regularlyusestwitterisverysmallcomparedtothevotingcommunityasawholeand
thosestudiesneedtoaccountforthatsomehow.Thethirdpointwassimplethatitneeds
tobeexplainable.

TweetsandVotes:AStudyofthe2011SingaporeGeneralElectionSkoric,M.,Poor,N.,
Achananuparp,P.,Lim,EP.,andJiang,J.2012.InProceedingsofthe45thHawaii
InternationalConferenceonSystemSciences.
TheauthorsrangefrombeingprofessorsinSingaporetoHongKongandtheirskillset
variesbut,overall,theyallareresearchesandstudiesofpublicopinionandtextdilution.

Thisstudy,likemanyothers,focusesonhowtwitterisabletopredictthepublic'sopinion
duringtheelectionseason.Itconcludedafterresearchthatthenamerecognitionworks
bestinthiscasescenarioascomparedtoreviewingthetweetsforthemeaningbehind
them.Themoreoftensomeonesnameismentionedduringanelection,thebettertheir
chancesareatwinning.However,thisstudyfoundthatitdependsonwhatthehottopics
areduringthatspecificelectionthatwilltargethowaccuratelysocialmediawebsitescan
predictsuccessorfailurerates.Whenthetopichassomethingtodowithsocialmedia,
peoplearemoreactiveinrespondingtoitonasiteliketwitterbecausemanyofthe
peopleontwitterarenewervotersandhaveopinionsonthatcomparedtoother
seeminglymoreboringtopics.Also,whenanelectioniscloseandcompetitionistight,

peoplearemoreactiveonsocialmedia,causingthelikelihoodofitbeingableto
successfullypredicttheoutcometogoup.

TweetingisBelieving?UnderstandingMicroblogCredibilityPerceptionsMorris,M.R.,Counts,
S.,Roseway,A.,Hoff,A.,andSchwarz,J.2012.InProceedingsofCSCW2012.
TheauthorsoftheirpieceareallprofessorsthatCarnegieMellonwheretheyworksinthe
fieldofHumanComputerinteractions.Thisarticlefocusesslightlylessonhowtheycan
predictelectionsandslightlymoreonhowmuchfaithpeoplehaveintwitterandthe
credibilityoftweets.Twitterinnaturesetsitselfuptheopportunityoffalsaccounts
projectingblatantlyincorrectinformationontothepublicanditisapparentlythejobof
thereadertofigureoutwhatittrueandwhatisnot.Peoplegenerallyarelesslikelyto
lookintotweetsaboutcelebritiesandtheirextravagantlylifeandmorelikelytoquestor
beskepticalofpoliticalinformationthattheyhaveyettohearfromanewssite.Theyare
alsomorelikelytobelievesomethingthatsomeonetheyfollowpostedratherthan
someonethatjustjustfoundthroughagoogleortwittersearch.Thereisacredibility
accountedwiththeamountofretweetsorfavoritesthatapersonoratweetreceives
however,accordingtothissurvey,thataccountslessinpeoplesmindsthanotherreasons
listedabove.

FromTweetstoPolls:LinkingTextSentimenttoPublicOpinionTimeSeriesOConnor,B.,
Balasubramanyan,R.,Routledge,B.R.,andSmith,N.A.2010.InProceedingsofthe
FourthInternationalAAAIConferenceonWeblogsandSocialMedia.

OConnorisaprofessoratUniversityofMassachusettsinAmherstwhereheworksinthe
computersocialsciencesfield.BalasubramanyanisagraduatestudentatCarnegie
Mellonwhereishestudyingmachinelearningandinformationextraction.Routledgeisa
professoratCarnegieMellonwhereheworksinthefinancedepartment.Smithisa
professorattheUniversityofWashingtonwherehestudiesandworksinthefieldof
ComputerSciencesandEngineering.

Intypicalcampaignsbeforethetimeofsocialmedia,pollingwastheonlyrealwaytoget
anaccuratefeelforthepublic'sopinion,butthisarticleprovesthatthatnolongerhasto
bethecase.BasedoninformationfromOpinionfinder,astudyofoveronebilliontweets
showsthatonecanadequatelygatherthepublic'sopinionquicklyand,forthemostpart,
accuratelybyusingkeywordstofindtweetstogagefeelings.Thebestwaytofindout
informationisbydoingthesesearchesassomethingimportantishappeningbecausethat
isthemajortimesthatpeoplearetweetingaboutit.Therearebillionsofresultsfollowing
anykeywordsearchandresearchersusedifferentwordstofindoutdifferentinformation:
Obamaforpresidentialapproval,EconomyandJobsforconsumerconfidence,etc.
Onedifficultyfoundinthisstudyisattemptingtotellthetonethatpeoplearewritingin
butthatwasfoundtobeanissuethatcouldbecombattedwiththerightformulas.Though
thisisnottrueinallofthecases,thestudyconcludedthat,withtherightresearchand
rightmeasures,twittercalculationscanbejustas,ifnotmoreeffective,thanpoles.

Fromobscuritytoprominenceinminutes:PoliticalspeechandrealtimesearchMustafaraj,E.,

andMetaxas,P.2010.InProceedingsofWebScience:ExtendingtheFrontiersofSociety
OnLine.
Thisarticlewasoriginallypublishedin2010bytwoprofessorsatWellesleyUniversity.
ThefirstisProfessorPanagiotisTakisMetaxas,aprofessorofcomputersciences
currentlyworkingmoreintothefieldofsocialmediaandandcrowdsurfingonsocial
platforms.ThesecondisProfessorEniMustafaraj,alsoaprofessorinthecomputer
sciencesfield.ShehighlightsonherwebsitethatasciencemagazinepublishedProfessor
MetaxasandherarticleentitledSocialMediaandtheElectionsin2012.

Thisisthefirstarticlethatreallytouchesonthetopicoftwitterbombing.Beforethe
popularityofsocialmediasitesbecamesovitaltothesuccessofanynewtopic(an
election,anewproduct,etc.),spamwasusedtogetamessageoutfastandeffectively
andtwitterbombingisessentiallythisgeneration'sversionofspam.Agroup,technically
unaffiliatedwithaspecificcampaign,createdtwitteraccountsthroughahackingwebsite.
Withinanhour,thisaccount,alongwiththeother10+madeatthattime,spamstwitter
userswithmessagesoflinksthatthatwouldnotnormallyfindthroughatypicalsearch
enginesfirstpage,whichveryfewpeoplelookbeyondthattogetanyinformationthey
need.Afterthis,thesuccessofthetwitterbombisdependantupontwothings:1)the
hopethattheytargetedtherightusersand2)thehopethatthoseusersretweetorpasson
thatmessageinanotherformat.

Thisdoesntalwaysworkout,butitisseenasagreatnew,technicallylegal,wasof
bombardingpeoplewithinformationthattheywouldnthavefoundontheirown.
Accordingtothestudy,thismethodworkssuccessfully,showingthatasingle
twitterbombedmessagecangetseenbyover61,000userswithinthefirstfewhoursof
itbeingposted.Therearenotreallydownsidestothisstrategy,butitssuccessisnot
alwaysguaranteed.Thiscouldbeviewedaspreachingtothechoir,meaningthatthe
groupofpeopletheysentthemessagetoeitherseeitandwouldhavealreadyseeniton
theirown,dontseeitbecausetheyarentactivelyontwitter,seeitbutdontretweetit,
orseeitandretweetittoanetworkthatwouldhaveseenitregardless.Regardlessofits
downsides,thisformofrelayinginformationgetsitouttothepublicquicklyandatno
cost,unlikeatelevisionorradioadvertisement.Thesetweettypicallyinvolveacallto
action,suchasaskingpeopletosignapetitionormaketheirvoiceheard,whichmakes
peoplewanttodosomething.

Sullivan,S.(2013,August14).HowTwittercouldpredictelectionsinoneeyecatching
study.
RetrievedOctober7,2015.
TheauthorofthisarticlebeganworkingatHotlineOnCall(NationalJournalsHotline
politicalblog)andatABCNewsandJapanPublicBroadcastingaftermajoringin
philosophyatHamiltonCollege.In2012,heswitchedfromthosejobstobeingareporter
atTheWashingtonPostcoveringtheinsandoutsofpoliticalelectionsandcampaigns.
Inthisarticle,heattemptstoexplainhow,whilesometimesthesocialmediapresence,

specificallytwitter,canpredictanelection,thatisntalwaysthecase.Onthelargescale,
thereisacorrelationbetweenthenumberoftweetsmentioningaspecificnameandtheir
likelihoodtowin,butthatdoesntworkforeverycandidateeverytime.Afterreading
over537milliontweetsinthemonthsleadinguptothepresidentialelection,theycome
totheconclusionthat,mostofthetime,anypublicityisgoodpublicity.

BasedonastudybyfourstudentsatIndianaUniversity,theamountacandidate'snameis
mentionedontwittersometimesrelatestohowlikelytheirchancesareofasuccessful
electionday.Thearticlereportsthatinrecentstudies,evennegativepublicityleadstoa
higherchanceofgettingelected.Inthepast,typicallytwitterresponsesfrommajor
politicaleventssuchasdebatestendtobemainlynegatively,seeminglyhurtinga
candidate'scampaignhowever,thisarticlestatesthatthepurenumberoftimestheir
namecomesup,inagoodorabadway,correspondsinapositivemannertothemgetting
elected.

Thisisnotalwaysthecase,though.Towardsthebottomofthearticle,thereisachartthat
showshowmucheachcandidatewastalkedaboutduringthe2012presidentialelection
andthesocialmediatalleyswerenotthesameaswhowaselected.RonPaul,acandidate
whodidnotmakeittobeingademocraticorrepublicannominee,wasbroughtupon
twittermoreoftenthanBarackObamaandMittRomneycombined.Hedidenduplosing
therepublicannomination,andthiscouldhavebeenforanumberofreasons,including
thathewasrunningasarepublicancandidatewithaprolife,antiwaragenda.Regardless

ofthereasonsthathelost,hisnameandopinionsweresopopularinthe2012election
thatsocialmediawouldhavepredictedthathewouldwintheentireelectionwithout
question.Thisshowsthat,eventhoughBarackObamaandMittRomneywerenumbers
twoandthreeonthelistonmentions,twitterandothersocialmediaplatformscannot
accuratelypredictelectionsonehundredpercentofthetime.

UnderstandingtheDemographicsofTwitterUsersMislove,A.,Lehmann,S.,Ahn,Y.Y.,
Onnela,
J.P.,andRosenquist,J.N.2011.InProceedingsoftheFifthInternationalAAAI
ConferenceonWeblogsandSocialMedia.
AlanMisloveisanassociateprofessoratNortheasternUniversityintheFieldof
ComputerandInformationSciences.

Socialmediasitesarepopularamongthegeneralpopulationand,becauseofthat,many
researchersattempttousethemasresearchtogagethepublic'sopinionhowever,most
sitesareprotectedbecauseofuserprivacybutTwitteristheexception,allowinganyone
accesstoover91%oftheprofilescreated.Beforeconductingresearch,onemust
understandthedemographicsofthegroupthatusesTwittertodetermineifitisan
accuraterepresentationofthepublic.Itwouldbeidealtocomparepropertiessuchas
socioeconomicclass,educationlevel,andemploymentstatus,but,becauseofuser
confidentiality,researcherscanonlyseethepersonsname,location,andthetweetsthat
theysend.Throughthosethreefactors,researcherscoulddeterminehowmuchofthe

populationgeographicallyitcanrepresentbasedontheU.S.Census,howagenderbias
mightcomeintoplay,andhowrace/ethnicityisimportant.Afterlookingintoallaspects
oftwitter,itbecameediantthatTwittermostlyrepresentsthetheU.S.comparedtoother
countries,malescomparedtofemales,andtherace/ethnicityofmostTwitterusersarenot
arandomsamplingofthepopulation.

DetectingandTrackingPoliticalAbuseinSocialMediaRatkiewicz,J.,Conover,M.D.,Meiss,
M.,Gonalves,B.,Flammini,A.,andMenczer,F.2011.InProceedingsoftheFifth
InternationalAAAIConferenceonWeblogsandSocialMedia.
ThisstudymainlytouchesonhowreliableorunreliableTwittercanbeandhowagood
politiciancouldusethistotheiradvantage.Socialmediaisallabouthowcatchy
somethingcanbesothatsomeonestopstopayattentiontoitwhilescrollingthrough
entirescreensofuselessinformation.Whilemanysourcesusetwittertogather
information,hoppingthatitisreliable,manypeopleareputtingoutblatantlyincorrect
statementsjustforthereaction.Unlikenewssites,therearenoanchors,writers,oreditors
onTwitter:noonetofactcheckinformationso,aswrongasitmayseem,themost
flashyversionofastorymakesheadlinesandmanypoliticianstakefulladvantageofthis.

OneofthewaysthatthisisdoneisthroughPoliticalAstroturfing.Astroturfingis
essentiallyattemptingtohidetheauthororpersonbehindanideatomakeitseemlikeit
iscomingfromthegrassrootsofanorganizationwhen,really,itiscomingfromastrong
andsuccessfulmarketingteam.ThisisdoneinthesamewaythataTwitterBombis:

creationofmanydifferentfakeusernamesquicklywithmassblastsofmessagestoas
manypeopleaspossible.Thiscanbeusedtohelpacandidatebygivingtheuserspositive
informationortohurtanothercandidatebyfeedingthemnegativityorsendingthemtoa
fakewebsite.Thisissimilartothewaythatspamissentviaemailhowever,whilethe
primaryobjectiveofaspammerisoftentopersuadeuserstoclickalink,someone
interestedinpromotinganastroturfmessagewantstoestablishafalsesenseofgroup
consensusaboutaparticularidea.Thisisusedtocreateafalseimpressionthatmore
peoplearetweetingandtalkingaboutacertaintopicthantheyactuallyare.

ThePartyIsOverHere:StructureandContentinthe2010Election
Livne,A.,Simmons,M.P.,Adar,E.,Adamic,L.A.2011.InProceedingsoftheFifth
InternationalAAAIConferenceonWeblogsandSocialMedia.
ThisarticletouchesonhowaccurateTwitteristodeterminethefutureorupcoming
successofacandidateandhowdifferentpoliticalpartiesusethiseithersuccessfullyor
not.Overall,afterdoingsubstantialresearchwithdifferentformulas,theyconcludedthat
twittercanbeusedtopredictelectionhowever,twittercannotbeusedtowinanelection.
Inthearticle,theycomparedittotheartofpoling.Thepollcantellareaderorresearcher
howthecampaignisgoingandhowthepublicisreactingtoit,butitwillnotchange
anyoneopinionsandthatisthesamewaytheyviewtwitter.Manypeoplewilluseitto
statetheiropinionsbutitwontchangeanythingexcepttotellreportersthatthisuseris
votingforthiscandidate.Also,thisarticleaddresseshowtheTeaPartyisvery
surprisinglysuccessfulatusingtwittertoadvancetheiragenda.Asitstates,theTeaParty

islargelyunorganizedanddoesnthavemuchofastructurehowever,theyareableto
verysuccessfullyusetwittertotheiradvantage.

OnUsingTwittertoMonitorPoliticalSentimentandPredictElectionResultsBermingham,A.,
andSmeaton,A.F.2011.InProceedingsoftheWorkshoponSentimentAnalysiswhere
AImeetsPsychology(SAAIP).
Thisarticlefirsttouchesonthreethingsthatpeopleareattemptingtoutilizetwittertoget
fromtheuser's:EventMonitoring,EmotiveSentiment,andElectionResults.Event
Monitoringisusedinrealtimeduringandshortlyaftersomethingimportanthappened
i.e.aspeech,adebate,etc.Twitterisusefulforthisbeforeresearchers,iftheyworkfast
enough,cangetresultsofthereactionsofthegeneralpopulation.Theymightevenbe
abletousethistochangehowaneventisgoingatthetimethatitishappening.Emotive
Sentimentisattemptingtofigureouthowpeopleviewsomethingthatisgoingoninless
ofapositivevs.negativewayandmoreofahowdoesitmakethemfeeltypeofway.
Peoplecancheckaboxonasurveyiftheresultofthisspecificincidentwaspositiveor
negativebutbyusing140characters,peoplegetrightdowntotheirpointofhowthey
feel.ThelastisElectionResultsandthisarticlestudiesthisinfourdifferentcategories:
timebased,samplesizebased,cumulative,andmanual.Overall,theyconcludedthat
volumeisamuchbetterandstrongerindicatorofanelectionsresultsthanthesentiment
describedwithinthetweets.

Diakopoulos,N.,&Shamma,D.(2010,April10).CharacterizingDebatePerformancevia

AggregatedTwitterSentiment.RetrievedOctober23,2015.
ThisarticledelvesintoTwitterbeforeitbecameextremelypopularandhowthemedia
firststartedtousesocialmediaasananalysistechnique.Thefirstdebatethatwastweeted
andshownonthetelevision,atopicwerefertonowaslivetweeting,wasthe2008
PresidentialDebatebetweenObamaandMcCain.Thisbecamesopopularthatevery
debateandlargeeventfollowingthiswasdonevialivetweeting.Thesuccessratewasby
theamountoftweetsthatcamethrough,butthereweresomefactorsthatwouldinfluence
theresults.Thewaythattheyknewwhowaslivetweetingwasthroughfollowingthe
hashtagsbutthereweremanytweetsthatwereaboutthedebatethatjustdidnthavethe
hashtaginitsotherewasnowaytotrackitatthetime.Thissystemwasalsobrokenin
thesensethatitcouldntaccountforthepeoplewatchingthedebatethatdidnthave
twitteraccountsorchoosenottolivetweetatthattime.

VocalMinorityversusSilentMajority:DiscoveringtheOpinionsoftheLongTailMustafaraj,
E.,
Finn,S.,Whitlock,C.,andMetaxas,P.T.2011.InProceedingsofPASSAT/SocialCom.
Thisarticleisthefirstonetotouchonthegroupswithinthetwittercommunitythatare
consideredthesilentmajority.Twitteristhefirstsocialmediaoutlettoallowpeopleto
holdaconversationwithothersinrealtime,whethertheyknowwhotheyretalkingtoor
not.Justlikereallife,therearesomepeoplewhodominateeveryconversationtheyarein
andotherswhositbackandaddinacommenthereorthere.Thevocalminorityis
consideredthesmallgroupofindividuals,orcompanies,thatdominatetheconversation,,

whetherotherpeopleagreewithwhattheyaresayingornot.TheSilentMajorityisthe
largergroupofpeoplewhojustsitbackandaddinonecommentbutoftenrepresent
moreofamajorityview,eventhoughitisjustaminorpartoftheconversation.

OneexampleofthiswasfromrightafterObamasfirstvictorywiththeYes,WeCan
campaign.Sincehegainedsuchafollowingonsocialmediaaccountsandendedup
victorious,peoplerightfullyassociatedthetwoofthemasdependantupononeanother
andcopiedthatsamemethod.Inasenateelection,twocandidateswerereallycloseand
thenonemadeahugesocialmediacampaignandpushformorefollowersandendedup
winning.Inthiscase,eventhoughthemajoritywassilent,thevocalminoritywasableto
createsuchabuzzandsuchafollowingthatitoversteppedtheamountofphysicalpeople
tweeting.

Massicotte,C.(n.d.).CampaigningonTwitter:140Charactersatatime.RetrievedOctober23,
2015.
Thisarticleislessofascholarlyarticlethanmanyoftheotheronesbutitstilladdresses
howimportanttwitterisinthepoliticalgame.Foundonwinningelections.org,asite
specificallyforwinningelections,thisarticleisallaboutthedosanddontswhenit
comestohavingatwitterforsomeonespersonalcampaign.Someofthedosaretobe
interestingandoriginal,tohavethecandidatetweet,notastaffmember,andtofollowas
manyrelevantpeopleasthecandidatecan.Someofthedontsaretonotfollowtoomany
peoplewhoarentimportanttothecampaign,dontlockthetwitter,anddonttweet

anythingthatthecandidatewouldbeembarrassedifitshoweduponthecoverpageofa
newspapertomorrowmorning.

InformationCredibilityonTwitterCastillo,C.,Mendoza,M.,andPoblete,B.2011.In
ProceedingsofWWW2011.
Thisarticlefocuseslessontheactualnotionofpoliticsbeingdiscussedontwitterand
howthataffectedtheelectionandmoreonhowmuchoftheinformationputoutis
valuableandcredible.ItalsoattemptstouncoverifusersonTwitterareabletotellwhat
informationshouldorshouldntbebelieved.Thefirststeptheresearcherstookwas
groupingtheinformationintotwocharatgoried:newsworthyandconversational.One
examplethattheygaveofnewsworthywastweetsthatmentionedthewordsearthday
comparedtotweetswiththekeywordhangoverwhichtendedtobemore
conversational.Thesetweetsarethenanalyzedfromafourprongedscope:message,user,
topic,andpropagation.Themessagelookedatwhatthephysicalaspectsofthetweet:
length,timeofdayitwassent,hasanymentions,hasbeenretweetedorwasaretweet,
etc.Theusershowedinformationaboutthepersonwhotweetedincludingtheirage,
demographicregion,iftheyareverified,iftheyhaveaurlintheirbio,etc.Thetopic
looksattheinformationofthetweet,includingquestionmarks,propernouns,some
specifichashtags,fractionofthetweetthatcontainsemoticons,etc.Thepropagation
looksatthemaxandminlevelsofpropagationusedinthetweet.Overall,theresearchers
cametotheconclusionthatthegeneralpubliccannotprocessinformationandjudgeits
credibilitythesamewayonlineastheycandoinreallife.


Predictingthe2011DutchSenateElectionResultswithTwitterTjong,E.,Sang,K.,andBos,J.
2012.InProceedingsofSASN2012,theEACL2012WorkshoponSemanticAnalysisin
SocialNetworks.
Thisarticlestartsoutalittlebitdifferentlyasitdoubtstheabilityforthepollsbeingused
currentlytocollecttherightdatabutexpressesthat,withalittlebitoftweaking,theright
formulacanbefoundandtheelectionpredictionpoweroftwittercanbejustasusefulas
normalpollingmeasures.ThisessentiallyrecreatestheexperimentdonebyTumasjanin
2010testingtheresultsofthe2009Germanelection.Theygatheredtweetsbysearching
forkeywordsandthennarrowingitdownbasedonlanguageandphrasing.Afterdoing
researching,theycameupwiththeconclusionthat,giventhecurrentsystemandtheir
slightlyalteredsystem,countingtweetscanassistinpredictionelectionshowever,they
werenotcompletelysatisfiedwiththeresults.Onereasonisbecausetheyplayedwiththe
numberstoaccountforthedifferenceindemographicsbetweenthetwitterusersandthe
actualvotersintheelection.Theyalsousedspecificnumberoftweetsasanindicators
but,associalmediaplatformsbecomemoreandmorepopular,thismethodofcounting
willnolongerworkbecauseallpartieswillhaveroughlythesameamountoftweets.

Boulianne,Shelley."SocialMediaUseandParticipation:AMetaanalysisofCurrentResearch,"
Information,Communication&Society,2015.
Thisarticleattemptstoanalyzetherelationshipbetweenthemediaandhowactivea
personisinthepoliticalspectrumbyreadingovervariousstudiesandcompiling

information.Manystudiesseemtoshowthatreadingwhatotherpeoplepostinfluence
peopletobecomemorepoliticallyengaged,regardlessofwhatsidetheinformation
postedcomesfrom.Fromthis,itgoesontosayhowclosefriendspostsmakeamuch
biggerdifferencethanpostsbystrangershowever,allpostsdochangepeoplesvoting
patterns.

Despitethis,thereisstillnotaguaranteedoragreeduponmethodofmonitoringthe
impactthatsocialmediareallyhas.Ofallofthedifferentstudiesconducted,theonethat
showsthebiggestlikelihoodthatthestudiescomedowntomorethanmerechanceisone
abouttheeffectsthatsocialmediahasonyouthvoters.Sincethereisnooneexperimental
design,researchesarestillveryunsureaboutthecausesandeffectsofsocialmediausage.
Forexample,politicalinterestcouldcausepeopletobecomeactiveonsocialmedia
platformsorbeingactiveonsocialmediaplatformscouldcauseaninterestinpolitics.

GayoAvello,D.Metaxas,P.T.andE.Mustafaraj.(2011)LimitsofElectoralPredictionsUsing
TwitterProceedingsoftheFifthInternationalAAAIConferenceonWeblogsandSocial
Media
Inthemedia,theuseofsocialmediahasbecomeincreasinglyimportanttopolitical
electionsfrompeopleonbothendsofthebeliefthattwittercanorcannotpredictthe
outcomeofvariouscampaigns.Ithasbecomeincreasinglyhypedup,butnotthroughany
reliablesourcespeoplemainlygettheirinformationfromusergeneratedblogs.The
numberofscholarlyarticlesaboutthisisinconsistentwiththeamountofmediacoverage,

leadingpeopletobelievemorethanwhatisproven.Afterresearching,itisrelativelyeasy
tofigureoutpublicapprovalviasocialmediaandthestudiesonthattendtobeaccurate
andconsistentwithpollingmeasures.Overall,thisarticle'sclaimisthatthereisnomore
likelihoodthattwittercanpredictanelectionthanchancehowever,withfutureresearch
andmoresetmeasuringtools,itisntimpossiblethattwittercouldbeusedwithaccuracy
soon.

GayoAvello,D.(2012,August28).No,YouCannotPredictElectionsWithTwitter(C.Petrie,
Ed.).RetrievedOctober25,2015.
Theauthorofthisarticlefirstbecameinterestedinthepredictivepowersoftheinternet
afterhearingagoogleemployeediscusshowgooglewasabletotellwhowouldwina
popularelectionbeforetheresultscameouttothepublicbasedonthenumberofgoogle
searches.Afterlookingintothis,theauthordiscoveredthatthisclaimwaswildly
unsupportedbecausethereweresomanyothervariablesthatcameintoplaythatwerenot
addressedwhensheexplainedhowtopredictelections.GayoAvellodecidedtocreate
hisownministudybasedaroundthistoprovehowlikelyitisthat,basedonsocial
media,anelectioncanbepredicted.

Afterdiscoveringhowdifficultusinggooglewas,heswitchedovertotwitterbutthiswas
noeasyfeateither.Usingthe2008U.S.PresidentialElection,hecollecteddatafromall
50statestodetermineiftheycouldhavepredictedtheresultsandfoundthat,yes,Obama
shouldhavebeenpredictedtowin,butnotintheslightmajoritythathedidfollowingthe

resultsfound,Obamashouldhavewoninalandslideineverysinglestatesbasedon
Twitterpredictions.Despitethisfinding,mostpeoplestillheavilybelievedinthe
predictivepowersoftwitterandpublishednumerousarticlesonhowamazingtheresults
were.Theauthorfoundthat,moreoftenthannot,papersarepublishedwithonlypositive
resultsbecausenegativeresultsoronesthatcontradictthegeneralpopulation'sopinion
arehardertobebelieved.Overall,theauthorfoundthatonecannotpredictelections
accuratelyeverytimeoffoftwitterremarksandsetssomeguidelinestowardstheendfor
howresearchersshouldtrytoimprovetheirownstudyonthissametopic.