AI-EnabledInfluenceOperations:SafeguardingFutureElections

SamStockwell,MeganHughes,PhilSwatton,AlbertZhang,JonathanHallandKieran

November2024

SamStockwell,MeganHughes,PhilSwatton,AlbertZhang,JonathanHallandKieran

AboutCETaS 2

Acknowledgements 2

ExecutiveSummary 3

CETaSUKelectionsecurityrecommendations 6

Introduction 10

Researchmethodology 11

Reportstructure 12

1.PublicVulnerabilityandResilienceagainstDeceptiveContent 13

1.1Riskfactorsassociatedwithvulnerability 13

1.2TheeffectsofAIonriskfactors 15

1.3Protectivefactorsassociatedwithresilience 17

2.AI-EnabledUSElectionThreatAnalysis 20

2.1QualitativeanalysisofAI-enabledUSelectionthreats 20

2.2NetworkanalysisofUSelectiondeepfakes 36

3.EvaluatingInfluenceOperationsintheAgeofAI 41

3.1Challengesinevaluatinginfluenceoperations 41

3.2Measuringhostileinfluenceoperations 43

4.PolicyResponsestoAI-EnabledElectionThreats 46

4.1Legalandregulatorymeasures 46

4.2Policymeasures 52

5.TechnicalSolutionstoAI-EnabledElectionThreats 59

5.1Preventionmethods 59

5.2Contentdetectionmethods 60

5.3Socialbotdetectionmethods 62

5.4Contentprovenance 63

Conclusion 66

AbouttheAuthors 67

1

AI-EnabledInfluenceOperations:SafeguardingFutureElections

AboutCETaS

TheCentreforEmergingTechnologyandSecurity(CETaS)isaresearchcentrebasedatTheAlanTuringInstitute,theUK’snationalinstitutefordatascienceandartificial

intelligence.TheCentre’smissionistoinformUKsecuritypolicythroughevidence-based,interdisciplinaryresearchonemergingtechnologyissues.ConnectwithCETaSat

cetas.turing.ac.uk.

ThisresearchwassupportedbyTheAlanTuringInstitute’sDefenceandNationalSecurityGrandChallenge.Allviewsexpressedinthisreportarethoseoftheauthors,anddonot

necessarilyrepresenttheviewsofTheAlanTuringInstituteoranyotherorganisation.

Acknowledgements

Theauthorsaregratefultoallthosewhotookpartinaworkshopforthisproject,withoutwhomtheresearchwouldnothavebeenpossible.Theauthorsarealsogratefulto:TonyAattheUK’sNationalCyberSecurityCentre;Anne-LouiseBrownattheAustralianCyber

SecurityCooperativeResearchCentre;DrJonathanBrightattheTuring’sPublicPolicy

Programme;researchersattheAustralianStrategicPolicyInstitute;andDanielJordan,

KevinXu,AliceCrillyandSamAbbottattheDepartmentforScience,Innovationand

Technologyforreviewinganearlierversionofthereport.ThefiguresinthisBriefingPaperweredesignedbyEmmaRowlandsandChrisRaggett.

ThisworkislicensedunderthetermsoftheCreativeCommonsAttributionLicence4.0,

whichpermitsunrestricteduseprovidedtheoriginalauthorsandsourcearecredited.Thelicenceisavailableat:

/licenses/by-nc-sa/4.0/legalcode.

Citethisworkas:SamStockwell,MeganHughes,PhilSwatton,AlbertZhang,JonathanHallandKieran,“AI-EnabledInfluenceOperations:SafeguardingFutureElections,”CETaS

ResearchReports(November2024).

2

SamStockwell,MeganHughes,PhilSwatton,AlbertZhang,JonathanHallandKieran

ExecutiveSummary

ThisCETaSResearchReportexamineshostileinfluenceoperationsenabledorenhancedbyartificialintelligence(AI),andmethodstoevaluateandcounteractsuchactivitiesduring

electioncyclesandbeyond.Italsoincludesevidence-basedanalysisofAI-enabledthreatsthatemergedintheNovember2024USpresidentialelection.

As2024drawstoaclose,morethan2billionpeopleinatleast50countrieswillhavevotedinthebiggestelectionyearinhistory.Atthestartoftheyear,thereweresignificant

concernsovertheproliferationofnewgenerativeAImodels,whichallowuserstocreateincreasinglyrealisticsyntheticcontent.Therehasbeenpersistentspeculationabouthow

thesetoolscoulddisruptkeyelectionsthisyear,manyofwhichwillhavemajorconsequencesforinternationalsecurity.

Therewasariskthatalackofempiricalworkontheimpactofthethreatwouldamplify

publicanxietyaboutit–which,inturn,couldhaveunderminedtrustinelectoralprocesses.Therefore,CETaScloselymonitoredkeyelectionsthroughouttheyear,tounderstandifandhowAImisuseaffectedtheseprocesses.AsreflectedintwoBriefingPaperspublishedin

MayandSeptember2024,CETaSconsistentlyfoundnoevidencethatAI-enabled

disinformationhadmeasurablyalteredanelectionresultinjurisdictionsrangingfromtheUKandtheEuropeanUniontoTaiwanandIndia.

ThisfinalResearchReportextendsthisglobalanalysistotheUSelectionandprovidesrecommendationsforprotectingtheintegrityoffuturedemocraticprocessesfromAI-enabledthreats,withafocusonhowUKinstitutionscancountersuchactivities.

KeyfindingsfromtheUSelectionspecificallyareasfollows:

?ThereisalackofevidencethatAI-enableddisinformationhashadameasurableimpactonthe2024USpresidentialelectionresults.However,thisisprimarilyduetoinsufficientdataontheimpactofsuchdisinformationonreal-worldvoter

behaviour.Whilesocialmediametricscanprovideinsightsintohowusersengage

withthiscontent,moreempiricalresearchisneededtounderstandhowitinfluenceslarge-scalevotingintentions.

?Despitethis,deceptiveAI-generatedcontentdidshapeUSelectiondiscoursebyamplifyingotherformsofdisinformationandinflamingpoliticaldebates.From

fabricatedcelebrityendorsementstoallegationsagainstimmigrants,viralAI-enabledcontentwasevenreferencedbysomepoliticalcandidatesandreceivedwidespreadmediacoverage.Nevertheless,non-AIfalsehoodscontinuedtohaveasignificant

3

AI-EnabledInfluenceOperations:SafeguardingFutureElections

impactandcouldnotbeignored.Theyincluded:misleadingclaimsbypolitical

candidates;conspiracytheoriespromotedbyfringeonlinegroups;andothertoolsofcontentmanipulation,suchastraditionalvideo-andimage-editingsoftware.

?AI-enableddisinformationintheUSelectionwasprimarilyendorsedoramplifiedbythosewithpre-existingbeliefsalignedwithitsmessages.Giventheextreme

politicalpolarisationofUSsociety,thecontentpredominantlyhelpedreinforcepriorideologicalaffiliationsamongtheelectorate.ThisechoespreviousCETaSfindingsthatalignmentbetweendisinformationandanindividual’sestablishedpolitical

opinionsiscrucialintheirdecisiontosharethecontent.

KeyfindingsforcounteractingAI-enabledinfluenceoperationsareasfollows:

?Digitalliteracy,astrongpublicbroadcastingecosystemandlowlevelsof

politicalpolarisationareallfactorsthatcanincreasepublicresistanceto

engagementwithdisinformation.Suchfactorspointtotheimportanceofinitiativestofosterahealthyinformationspaceatboththeindividualandsocietallevels.

?Thereisnoone-size-fits-allframeworktoevaluatehostileinfluenceoperationstargetingfutureelectioncyclesorwidersociety.Instead,researchersshould

weighthetrade-offsbetweenthedifferenttoolsthatareavailableandusetheonemostsuitedtotheoperationinquestion.Insomecases,combiningdifferent

frameworkswillprovideadditionalinsightsintotheseactivities.

?GiventhesignsthatAI-enabledthreatsbegantodamagethehealthof

democraticsystemsgloballythisyear,complacencymustnotcreepinto

governmentdecision-making.Aheadofupcominglocal,regionalandnational

elections–fromAustraliaandCanadain2025toScotlandin2026–thereisnowa

valuableopportunitytoreflectontheevidencebaseandidentifymeasurestoprotectvoters.

?Therefore,thisreportrecommendsthefollowingactionstoprotectelections

andwidersocietyfromAI-enabledinfluenceoperationsandotherdisinformationactivities.Thesesolutionshavebeeninformedbyanextensiveliteraturereviewandworkshopswith47cross-sectorexperts.TheycentreonthefollowingfourstrategicobjectivesdesignedtohelpUKinstitutionstargetdifferentaspectsoftheonline

disinformationprocess:

oCurtailgeneration–measuresthatincreasebarriersto,ordeteractorsfrom,creatingonlinedisinformationinthefirstplace.

oConstraindissemination–measuresthatreducetheeffectivenessandviralityofdisinformationcirculatingondigitalplatforms.

4

SamStockwell,MeganHughes,PhilSwatton,AlbertZhang,JonathanHallandKieran

oCounteractengagement–measuresthattargetthewaysthatusers

consumedisinformationondigitalplatforms,toreducemaliciousinfluence.

oEmpowersociety–measuresthatstrengthensocietalcapabilitiesforexposingandunderminingonlinedisinformation.

5

AI-EnabledInfluenceOperations:SafeguardingFutureElections

CETaSUKelectionsecurityrecommendations

Curtailgeneration

1)DigitalprovenancestrategyforUKorganisations:TheUKDepartmentfor

Science,InnovationandTechnology(DSIT)shouldestablishanimplementation

strategyforautomaticallyembeddingprovenancerecordsindigitalcontent

producedbytheUKGovernmentandothersectorsatitsorigin.Thiswould

strengthentheauthenticityofcredibleinformationsources,andcoulddrawontheUSOfficeofManagementandBudget’srequirementtoissuesimilarguidancebyJune2025.

2)Authenticity-by-design:TheInternetEngineeringTaskForce’sSecurityAreashoulddevelopandimplementauthenticity-by-designprinciplesacrossthe

internetecosystemtoprotectinformationintegrity,usingstructuressuchastheStarlingLabframework.Theschemeshouldaimtoembedtoolsintodifferent

partsoftheinternetinfrastructurethatautomaticallycapture,storeandverifydigitalprovenancerecordssecurely.

3)Clarifyingexistinglaws:TheUKMinistryofJusticeshouldconductareviewtounderstandweaknessesinexistinglegislationthatmaybeexploitedwith

maliciousAI-generatedcontenttargetingpoliticalcandidatesordesignedto

undermineelectionintegrity(includingthoserelatedtodefamation,privacyandelectorallaws).ThiswillhelptheMinistryunderstandwhetherexistinglawsareadequatetodetersuchactivitiesorwhetherlegislativereformsarerequired.

Constraindissemination

4)Deepfakedetectionbenchmarkingandguidance:TheUKAISafetyInstituteandtheHomeOfficeshouldcoordinatetodevelopstandardisedbenchmarksand

guidancefordeepfakedetectiontools,providingminimumqualityassurancesfor

6

SamStockwell,MeganHughes,PhilSwatton,AlbertZhang,JonathanHallandKieran

thoseusingthem.Thebenchmarksshouldbecontinuouslyupdatedagainstnewdeepfakeexamplestomaintainrelevance,whiletheguidanceshouldencouragedeveloperstopublishalistofkeydetailsbeforerelease,including:thepurposeandscopeofthetool;howitshouldbeusedandinterpreted;theexplainabilityofitsoutputs;anditslimitations.

5)Codeofconductondisinformation:AspartofitsPhaseThreeroadmapfortheOnlineSafetyAct2023,OfcomshouldcreateanewCodeofConductaimedatsystematicallytargetingonlinedisinformation.DrawinginspirationfromtheEU’sCodeofPracticeonDisinformation,thenewcodeshouldsetoutself-regulatorystandardsfordifferentsectorsondemonetisingdisinformationcontentcreators;defineunpermittedmanipulativebehavioursassociatedwithbotaccounts;

providetoolsforempoweringusersagainstdisinformation;andrequiretransparentincidentreporting.

6)Politicalpartyconduct:TheElectoralCommissionshouldexpandexisting

guidanceforUKpoliticalpartiesonboththeappropriateuseofAItoolsandclearredlinesonmisuse.Inturn,politicalpartiesshouldupdatetheirinternalcodesofconductwiththisguidancetocreateaccountabilityforcandidatesand

campaigners.

Counteractengagement

7)Mediavalidationapptools:OfcomshouldconvenemajorUKcommunicationsappprovidersandtheInternationalFact-CheckingNetworktodesignaccessibleandtransparentfact-checkingappsforUKusers.Thesecouldreplicateother

initiatives,suchasTaiwan’sLINEapp,whichhelpsusersverifycontentbyprovidingtrustedalternativenewssourcesforcross-referencing.

7

AI-EnabledInfluenceOperations:SafeguardingFutureElections

8)ElectionIncidentProtocol:TheCabinetOfficeshouldestablishaUKCritical

ElectionIncidentPublicProtocolbasedontheCanadianmodel.Involvingarangeofseniorgovernmentexperts,theprotocolwouldinformthepublicofthreats

consideredsevereenoughtounderminetheintegrityofelections.Any

announcementsmadethroughtheprotocolshouldbebasedonaconsensusandrestrictedtoinformingthepublicabouttheincidentandhowtheycanprotect

themselves.

9)Electionadvertimprints:TheUKGovernmentshouldtableanamendmenttoSection54oftheElectionsAct2022,whichdealswithimprintsondigital

campaignmaterialduringelections.Thisshouldintroduceanewtransparencyprovisionlegallyrequiringadvertcontentthathasbeendigitallyeditedtobe

embeddedwithsecureprovenancerecordsdetailinghowitwaseditedandbywhom.

10)Decentralisedfact-checking:Socialmediaplatformsshouldinvestgreater

resourcesinsupportofdecentralisedfact-checkinginitiatives,tohelpaddressthevolumeofdisinformationcirculatedonline.Theseinitiativesshouldincorporate

reputationandvotingsystemstoprovidequalitycontrolof,andademocraticconsensuson,user-madenotices.

11)Mediareportingguidance:TheIndependentPressStandardsOrganisation

shouldreviseitsexistingguidanceon‘reportingmajorincidents’toincludekey

considerationsforcoverageofknownhostileinfluenceoperationsandviral

disinformationcontent–drawingoninsightsfromjournalistsandfact-checkers.Suchinformationcouldincludeadvicetorefrainfromlinkingtotheoriginalsourcecontentinonlinearticles–therebydiscouragingusersfromsharingitwithothers–andtoframetheimpactofthecontentinawaythatdoesnotexaggeratethe

threatoftheseactivitiestothewiderpublic.

Empowersociety

12)Regulatorreview:DSIT’sAICentralRiskFunctionshouldcoordinatewithboththeElectoralCommissionandOfcomtoanalysepotentialgapsintheirrespective

8

SamStockwell,MeganHughes,PhilSwatton,AlbertZhang,JonathanHallandKieran

regulatorypowersandremits.Thereviewshouldfocusontheeffectivenessofbothregulatorsintacklingallformsofonlinedisinformationduringelections,inaccordancewiththeOnlineSafetyAct2023,theElectionsAct2022andthe

RepresentationofthePeopleAct1983.

13)Trustedresearcheraccess:TheUKGovernmentshouldensurethattheDigitalInformationandSmartDataBillandotherrelevantfuturelegislationincludea

provisionforestablishingatrustedresearchgroupondisinformation.ThiswouldrequiresocialmediaplatformstoprovidetrustedmembersoftheUKacademic,researchandcivilsocietycommunitieswithaccesstodataonidentifiedhostile

influenceoperations–akintoX’sformerdataaccessmodel.Tomaintain

impartiality,organisationsandindividualsshouldbeselectedbyUKResearchandInnovation’strustedresearchandinnovationprogramme.

14)Conveningexperts:OfcomshouldprioritiseestablishingtheAdvisoryCommitteeonDisinformationandMisinformation,assetoutbysection152oftheOnline

SafetyAct2023,tomaintainalong-termfocusontacklingdisinformation.The

committeeshouldhaveaclearmandateforinformingOfcom’scounter-

disinformationactivities,anindependentchairnotaffiliatedwithanypoliticalpartyortechplatform,anddiversesectoralrepresentation.

15)Digitalliteracyprogrammes:TheDepartmentforEducationandDSITshould

coordinateonestablishingnationwidedigital-literacyandcritical-thinking

programmes.Anyschemesofthiskindshouldbemademandatoryinprimaryandsecondaryschools,whilealsobeingpromotedtoadults.Suchinitiativeswould

seektoimprovesocietalresilienceagainstdisinformationandcouldincludetopicson:AIandalgorithmicbias;deepfakes;evaluatinginformationsources;understandingsocialmediamanipulation;andbuildingacultureofcontentverification.

9

AI-EnabledInfluenceOperations:SafeguardingFutureElections

Introduction

SinceCETaSpublisheditsBriefingPaperontheUK,EUandFrenchelectionsinSeptember2024,mostvotingprocesseshaveconcludedwithoutbeingfundamentallyreshapedor

disruptedbyAI.However,atthetimeofwriting,thepivotal2024USpresidentialelection

hadnottakenplace.Giventhelongtimeframeofthecampaign,itsnarrowpollmarginsandthedifferencesbetweenthetwomaincandidatesonRussiaandChinapolicy,many

observersbelievedtheelectionwouldbetheultimatetestofAI-generateddisinformation.1

YetaspreviousCETaSresearchconcluded,thereisaneedtoinformsuchjudgementswithevidence-basedresearchandfindabalancebetweenassessingtheseverityofthethreat

andavoidingfearmongering.2AIthreatreportinginthecontesthasfocusedonunpickingindividualviralcasesinsteadofsystematicanalysisofstrategicthemesandtrendsacrosstheelectioncycle.Onlywell-groundedresearchcanaccuratelyinformthepublicandavoidunnecessaryspeculation.

The‘superyearofelections’maybedrawingtoaclosebutAImisusecouldstillemergeinfederalelectionsinAustraliaandCanadain2025,aswellasinregionalelectionssuchasthoseinScotlandin2026.Thereisariskthateffortstotacklethesethreatswillbe

deprioritisedontheincorrectassumptionthat,withmanynationalelectionsnowfinished,maliciousactorswillhavelittleincentiveforfurtherpoliticalinterference.Butmaintainingahealthyinformationenvironmentisalsocrucialoutsideelectionperiods,asevidencedintheUKcontextbytherecentuseofdisinformationtointensifyfar-rightriotsandpolitical

extremism.3

Therefore,policyresponsesandotherprotectivemeasuresshouldnotbenarrowlyfocusedonsecuringelectioncyclesonlyasofficialcampaigningtakesplace.4Instead,theyshouldidentifylong-terminterventionsthatembedresilience,drawonthecapabilitiesofdifferent

1WilliamTurton,“TheUSElectionThreatsAreClear.WhattoDoAboutThemIsAnythingBut,”WIRED,15May2024,

/story/election-threats-senate-hearing-ai-disinformation-deepfakes/.

2SamStockwelletal.,“AI-EnabledInfluenceOperations:TheThreattotheUKGeneralElection,”CETaSBriefingPapers(May2024),39

,https://cetas.turing.ac.uk/publications/ai-enabled-influence-operations-threat-uk-general-election;

SamStockwell,“AI-EnabledInfluenceOperations:ThreatAnalysisofUKandEuropeanElections,”CETaSBriefingPapers(September2024),6.

3InstituteforStrategicDialogue,“Fromrumourstoriots:Howonlinemisinformationfuelledviolenceintheaftermathofthe

Southportattack,”31

July2024,/digital_dispatches/from-rumours-to-riots-how-online-

misinformation-fuelled-violence-in-the-aftermath-of-the-southport-attack/.

4HelenMargetts,“TheAlelectionthatwasn’t–yet,”UKElectionAnalysis

,https://www.electionanalysis.uk/uk-election-

analysis-2024/section-6-the-digital-campaign/the-al-election-that-wasnt-yet/.

10

SamStockwell,MeganHughes,PhilSwatton,AlbertZhang,JonathanHallandKieran

sectorsandempowercitizensagainstdisinformation.Allsuchstepswillhelpprotectfutureelections–andwidersociety–againstthesethreats.

Researchmethodology

Thisprojectseekstoanswerthefollowingresearchquestions:

?RQ1:Whatfactorsmakeindividualcitizensandsocietyeithermorevulnerableorresilienttoengagementwithdisinformation,includingAI-enabledcontent?

?RQ2:HowhasAIbeenmaliciouslydeployedinthelead-uptothe2024USpresidentialelection?

?RQ3:Whichexistingevaluationframeworksgaugetheimpactofinfluenceoperations,andwhatarethebarrierstoeffectivemeasurement?

?RQ4:WhatinitiativescantheUKimplementtoenhanceelectionsecurityand

broadersocietalresilienceagainstinfluenceoperationsthatincorporatenovelAItools?

DatacollectionforthisstudywasconductedbetweenJuneandNovember2024,involvingthreecoreresearchactivities:

1.Literaturereviewcoveringjournalarticles,publicreportsandnewsarticleson:AImisuseinthe2024USelection;publicengagementwithAI-enableddisinformation;challengesinevaluatinginfluenceoperations;andcountermeasuresforimprovingelectionresilience.

2.SocialmediaanalysisofthreedifferentUSdeepfakes,tounderstandnodesofinfluenceamplifyingdisinformation(seeSection2.2formoredetailsonthe

methodologyused).

3.Twoworkshopsdesignedtoprioritisepolicyandtechnicalrecommendations

identifiedbytheprojectteam.ThesesessionsinvitedattendeestodeterminewhichsolutionsweremostimpactfulandfeasibleinenhancingelectionresilienceagainstAIthreats;theyinvolved47experts:

?20fromindustry.

?12fromgovernmentandregulatorybodies.

11

AI-EnabledInfluenceOperations:SafeguardingFutureElections

?10fromcivilsociety.

?5fromacademia.

Reportstructure

Theremainderofthisreportisstructuredasfollows.Section1describesthefactorsthataffectindividualandsocietalengagementwithdisinformationcontent.Section2providesanalysisofspecificAIthreatsinthe2024USpresidentialelectioncycle,aswellassocialmediaanalysisofaselectionofhigh-profileUSdeepfakes.Section3exploreschallengesandwaysforwardinevaluatingtheimpactoftheseactivities.Section4describespolicysolutionsthatcanhelpincreasedemocracies’resilienceagainstmaliciousAI-enabled

influenceoperations.Finally,Section5detailscorrespondingtechnicalsolutions.

12

SamStockwell,MeganHughes,PhilSwatton,AlbertZhang,JonathanHallandKieran

1.PublicVulnerabilityandResilienceagainstDeceptiveContent

Theubiquityofsocialmediahasshiftedresponsibilityfordetectingfalsehoodsfrom

professionaljournaliststoeverydayinternetusers.Itis,therefore,importanttounderstandhowindividualsinteractwithmisinformationanddisinformationtoreducepublic

susceptibilitytoit–bothduringelectionsandbeyond.CETaSdefinesmisinformationas

unintentionallymisleadingclaims.Incontrast,disinformationisdeliberatefalsehoods,

includingthosesharedaspartofonlineinfluenceoperationsthatareintendedtoshape

publicopinionorbehaviour.Theanalysisinthissectionfocusesontheriskfactorsand

protectivefactorsthataffectpublicvulnerabilityandresilienceagainstbothmisinformationanddisinformation.

1.1Riskfactorsassociatedwithvulnerability

Tounderstandtheimpactofmisinformationanddisinformationonindividuals,itishelpfultobreakdownthedifferentstagesofthecontentlifecycle:

Figure1.Onlinemisinformationanddisinformationlifecycle

Source:Authors’analysis.

13

AI-EnabledInfluenceOperations:SafeguardingFutureElections

Thereisonlysparsedataontheknownmotivationsofpeoplewhogeneratemisinformationanddisinformation.5However,somefactorshavebeensuggestedbasedonhistoricalcases.Theseincludevariousforeignanddomesticactors’desiretoinfluenceelectionresults,sowpoliticaldivisionorunderminemediaintegrityinacountry,aswellashyper-partisanmediaoutlets’aimtodistortfactstosuitorganisationalagendas.6

Anincreasingbodyofevidencehelpsexplainthereasonswhyindividualsdisseminate,

engageandpositivelyreacttothiscontent.Forexample,individualswhoconsume

misinformationanddisinformationaremorelikelytohaveconspiratorialoutlooks,distrust

publicinstitutions,experiencestressandfrustration,orlackcritical-thinkingand

information-verificationhabits.7Individualswhoexaggeratetheirknowledgeoftopicsandscorelowerontestsofanalyticalthinkingarealsomorelikelytobelievefakenewsstories.8Whenitcomestodemographics,somestudiesshowthatolderpeoplearemorelikelyto

sharemisinformationordisinformationonlinewhentheyviewit,butyoungerpeople–

particularlythoseundertheageofeighteen–aremorelikelytobelievemisleading

narratives.9Otherstudieshavefoundthatmenaremorelikelythanwomentodisseminatepoliticaldisinformation.

Userswhorelyonsocialmedia(ratherthantraditionalmedia)fornewsandpolitical

engagementwillalsobemorelikelytoencountermisinformationanddisinformation–andwill,therefore,bemoreatriskofconsumingit.Disinformationmaybespreadbygroupswith

ideologicalagendas,suchasclimate-changedeniers,orbythoseseekingtobenefit

5SophieLechelerandJanaLauraEgelhofer,“Disinformation,Misinformation,andFakeNewsUnderstandingtheSupplySide”inKnowledgeResistanceinHigh-ChoiceInformationEnvironments,ed.JesperStr?mb?cketal.(Routledge:2022),73-80,

/bitstream/handle/20.500.12657/54482/1/9781000599121.pdf.

6Ibid.

7ValentinStoian-IordacheandIrenaChiru,”2.AggravatingFactorsfortheDisseminationofDisinformation:2.1.Individualandgroupfactors”inHandbookonIdentifyingandCounteringDisinformation,ed.ChristinaIvanetal.(DOMINOESProject:2023),

https://dominoes.ciberimaginario.es/21-individual-factors.html;

Joná?Syrovátka,NikolaHo?ej?andSarahKomasová,

“Towardsamodelthatmeasurestheimpactofdisinformationonelections,”EuropeanView22,no.1(2023),

/10.1177/17816858231162677.

8GordonPennycookandDavidG.Rand,“Whofallsforfakenews?Therolesofbullshitreceptivity,overclaiming,familiarity,andanalyticthinking,”JournalofPersonality88,No.2(March2019:185-200)

,/10.1111/jopy.12476.

9AndrewGuess,JonathanNaglerandJoshuaTucker,“Lessthanyouthink:Prevalenceandpredictorsoffakenews

disseminationonFacebook,”ScienceAdvances5,no.1(January2019);CenterforCounteringDigitalHate,“Beliefin

conspiracytheorieshigheramongteenagersthanadults,asmajorityofAmericanssupportsocialmediareform,newpolling

finds,”

16August2023,/doi/10.1126/sciadv.aau4586;/blog/belief-in-

conspiracy-theories-higher-among-teenagers-than-adults-as-majority-of-americans-support-social-medi