Scholars' Mine
Scholars' Mine
Business and Information Technology FacultyResearch & Creative Works Business and Information Technology
01 Apr 2020
Artificial Intelligence (AI) Ethics: Ethics of AI and Ethical AI
Artificial Intelligence (AI) Ethics: Ethics of AI and Ethical AI
Keng Siau
Missouri University of Science and Technology, siauk@mst.edu Weiyu Wang
Follow this and additional works at: https://scholarsmine.mst.edu/bio_inftec_facwork
Part of the Business Commons
Recommended Citation Recommended Citation
Siau, K., & Wang, W. (2020). Artificial Intelligence (AI) Ethics: Ethics of AI and Ethical AI. Journal of Database Management, 31(2), pp. 74-87. IGI Global.
The definitive version is available at https://doi.org/10.4018/JDM.2020040105
This Article - Journal is brought to you for free and open access by Scholars' Mine. It has been accepted for inclusion in Business and Information Technology Faculty Research & Creative Works by an authorized administrator of Scholars' Mine. This work is protected by U. S. Copyright Law. Unauthorized use including reproduction for redistribution requires the permission of the copyright holder. For more information, please contact scholarsmine@mst.edu.
DOI: 10.4018/JDM.2020040105
Copyright©2020,IGIGlobal.CopyingordistributinginprintorelectronicformswithoutwrittenpermissionofIGIGlobalisprohibited.
Artificial Intelligence (AI) Ethics:
Ethics of AI and Ethical AI
Keng Siau, Missouri University of Science and Technology, Rolla, USA Weiyu Wang, Missouri University of Science and Technology, Rolla, USA
ABSTRACT Artificialintelligence(AI)-basedtechnologyhasachievedmanygreatthings,suchasfacialrecognition, medicaldiagnosis,andself-drivingcars.AIpromisesenormousbenefitsforeconomicgrowth, socialdevelopment,aswellashumanwell-beingandsafetyimprovement.However,thelow-levelof explainability,databiases,datasecurity,dataprivacy,andethicalproblemsofAI-basedtechnology posesignificantrisksforusers,developers,humanity,andsocieties.AsAIadvances,onecritical issueishowtoaddresstheethicalandmoralchallengesassociatedwithAI.Eventhoughtheconcept of“machineethics”wasproposedaround2006,AIethicsisstillintheinfancystage.AIethicsis thefieldrelatedtothestudyofethicalissuesinAI.ToaddressAIethics,oneneedstoconsiderthe ethicsofAIandhowtobuildethicalAI.EthicsofAIstudiestheethicalprinciples,rules,guidelines, policies,andregulationsthatarerelatedtoAI.EthicalAIisanAIthatperformsandbehavesethically. OnemustrecognizeandunderstandthepotentialethicalandmoralissuesthatmaybecausedbyAI toformulatethenecessaryethicalprinciples,rules,guidelines,policies,andregulationsforAI(i.e., EthicsofAI).WiththeappropriateethicsofAI,onecanthenbuildAIthatexhibitsethicalbehavior (i.e.,EthicalAI).ThispaperwilldiscussAIethicsbylookingattheethicsofAIandethicalAI. WhataretheperceivedethicalandmoralissueswithAI?Whatarethegeneralandcommonethical principles,rules,guidelines,policies,andregulationsthatcanresolveoratleastattenuatetheseethical andmoralissueswithAI?Whataresomeofthenecessaryfeaturesandcharacteristicsofanethical AI?HowtoadheretotheethicsofAItobuildethicalAI? KEywoRDS
AI Ethics, Artificial Intelligence, Ethical AI, Ethics, Ethics of AI, Machine Ethics, Roboethics
1. INTRoDUCTIoN Someresearchersandpractitionersbelievethatartificialintelligence(AI)isstillalongwayfrom havingconsciousnessandbeingcomparabletohumans,andconsequently,thereisnorushtoconsider ethicalissues.ButAI,combinedwithothersmarttechnologiessuchasrobotics,hasalreadyshown itspotentialinbusiness,healthcare,transportation,andmanyotherdomains.Further,AIapplications arealreadyimpactinghumanityandsociety.Autonomousvehiclescanreplacealargenumberof jobs,andtransformthetransportationandassociatedindustries.Forexample,short-haulflightsand
hospitalityservicesalonghighwayswillbeimpactedifdriverlesscarsenablepassengerstosleep andworkduringthejourney.AI-recruitersareknowntoexhibithumanbiasesbecausethetraining datainheritsthesamebiaseswehaveashumans.Thewealthgapcreatedbythewideningdifferences betweenreturnoncapitalandreturnonlaborisposedtocreatesocialunrestandupheavals.The futureofworkandfutureofhumanitywillbeaffectedbyAIandplansneedtobeformulatedandput inplace.BuildingAIethicallyandhavingethicalAIareurgentandcritical.Unfortunately,building ethicalAIisanenormouslycomplexandchallengingtask. 1.1. what is Ethics? Ethicsisacomplex,complicated,andconvolutedconcept.Ethicscanbedefinedasthemoralprinciples governingthebehaviorsoractionsofanindividualoragroupofindividuals(Nalini,2019).Inother words,ethicsareasystemofprinciplesorrulesorguidelinesthathelpdeterminewhatisgoodor right.Broadlyspeaking,ethicscanbedefinedasthedisciplinedealingwithrightversuswrong,and themoralobligationsanddutiesofentities(e.g.,humans,intelligentrobots,etc.). Ethicshasbeenstudiedbymanyresearchersfromdifferentdisciplines.Mosthumansarefamiliar withvirtueethicssinceveryyoungbecauseitisabehaviorguideinstilledbyparentsandteachersto helpchildrenpracticegoodconduct.Aristotle(Yu,1998)believeswhenapersonactsinaccordance withvirtue,thispersonwilldowellandbecontent.Virtueethicsispartofnormativeethics,which studieswhatmakesactionsrightorwrong.Itcanbeviewedasoverarchingmoralprinciplesthathelp peopleresolvedifficultmoraldecisions.Astheinteractionbetweenhumans,betweenhumansand animals,betweenhumansandmachines,andevenbetweenmachinesisincreasing,ethicaltheories havebeenappliedtoreal-lifesituations,suchasbusinessethics,animalethics,militaryethics, bioethics,andmachineethics.Thestudyofethicsandethicalprinciplesisconstantlyevolvingand developing.Table1listsseveralethicsdefinitionsgivenbyresearchers. InthecontextofAI,theethicsofAIspecifiesthemoralobligationsanddutiesofanAIandits creators.Researchershavedonemuchworkstudyinghumanethicalissues.Manyethicalframeworks canbeusedtodirecthumanbehaviors,suchasactionsandactivitiesrelatedtorespectforindividuals,
Table 1. Definition of ethics
Normative Ethics Reference
Ethicsisthecapacitytothinkcriticallyaboutmoralvaluesanddirectouractionsin termsofsuchvalues. Churchill,1999 Ethicsisasetofconceptsandprinciplesthatguideusindeterminingwhatbehavior helpsorharmssentientcreatures. Paul&Elder,2006 Ethicsisthenormforconductthatdistinguishesbetweenacceptableand unacceptablebehavior. Ethicsisthedisciplinethatstudiesstandardsofconduct,suchasphilosophy, theology,law,psychology,orsociology. Ethicsisamethod,procedure,orperspectivefordecidinghowtoactandfor analyzingcomplexproblemsandissues. Resnik,2011 Applied Ethics Computerethicsistheanalysisofthenatureandsocialimpactofcomputer technologyandthecorrespondingformulationandjustificationofpoliciesforthe ethicaluseofsuchtechnology. Moor,1985,p.266 Machineethicsisconcernedwithgivingmachinesethicalprinciplesoraprocedure fordiscoveringawaytoresolvetheethicaldilemmastheymightencounter,enabling themtofunctioninanethicallyresponsiblemannerthroughtheirownethical decisionmaking. AndersonandAnderson,2011, p.1
beneficence,justice,privacy,accuracy,ownership/property,accessibility,fairness,accountability, andtransparency(Wang&Siau,2018). Oneofthebest-knownethicalframeworksisdevelopedbyKenBlanchardandNormanVincent Peale(Blanchard&Peale,2011).Theframeworkconsistsofthreemainquestions:Isitlegal?Is itfair?Howdoesitmakemefeel?AnotherframeworkistheMarkkulaCenterFramework,which identifiesfiveapproachestodealingwithethicalissues,includingtheutilitarianismapproach,rights approach,fairnessorjusticeapproach,commongoodapproach,andvirtueapproach(Markkula CenterforAppliedEthics,2015). AIethics,however,isarelativelynewfieldandthesubsequentpartsofthepaperwilldiscuss theAIethics–ethicsofAIandethicalAI.
1.2. what is “Ethics of AI” and “Ethical AI”?
TheethicsofAIispartoftheethicsofadvancedtechnologythatfocusesonrobotsandotherartificially intelligentagents.Itcanbedividedintoroboethics(robotethics)andmachineethics. Roboethicsisconcernedwiththemoralbehaviorsofhumansastheydesign,construct,use,and interactwithAIagents,andtheassociatedimpactsofrobotsonhumanityandsociety.Inthispaper, weconsideritasethicsofAI,whichdealwithethicalissuesrelatedtoAI,includingethicalissues thatmayarisewhendesigninganddevelopingAI(e.g.,humanbiasesthatexistindata,dataprivacy, andtransparency),andethicalissuescausedbyAI(e.g.,unemploymentandwealthdistribution). Further,asmachinesbecomemoreintelligentandmayonedaygainconsciousness,weshouldconsider robotrights--theconceptthatpeopleshouldhavemoralobligationstowardsintelligentmachines.It issimilartohumanrightsandanimalrights.Forinstance,whetheritisethicaltodeployintelligent militaryrobotstodangerousbattlefieldsorassignrobotstodirtyenvironments.Therightsofliberty, freedomofexpression,equality,andhavingthoughtandemotionbelongtothiscategory. MachineethicsdealswiththemoralbehaviorsofArtificialMoralAgents(AMAs),whichis thefieldofresearchaddressingthedesignofartificialmoralagents.Astechnologyadvancesand robotsbecomemoreintelligent,robotsorartificiallyintelligentagentsshouldbehavemorallyand exhibitmoralvalues.WeconsidertheethicalbehaviorsofAIagentsasethicalAI.Currently,the bestknownproposedrulesforgoverningAIagentsaretheThreeLawsofRoboticsputforthbyIssac Asimovinthe1950s(Asimov,1950).FirstLaw,arobotmaynotinjureahumanbeingor,through inaction,allowahumanbeingtocometoharm.SecondLaw,arobotmustobeytheordersgivento itbyhumanbeingsexceptwhensuchorderswouldconflictwiththeFirstLaw.ThirdLaw,arobot mustprotectitsexistenceaslongassuchprotectiondoesnotconflictwiththeFirstorSecondLaw. Table2depictsthetwodimensionsofAIethics(i.e.,ethicsofAIandethicalAI)andhowthe twodimensionsinteractwithAI,Human,andSociety.TheethicalinteractionbetweenAIsisnew inthispaper.ThisisespeciallyimportantforAIswithconsciousness.NotonlyshouldtheAIsdo noharmtohumansandself-preserve,butitalsoshoulddonoharmtootherintelligentagents. Thus,theThreeLawsofRoboticsmayneedtobeextendedtotakeintoaccounttheinteraction betweenintelligentAIs. UnderstandingtheethicsofAIwillhelptoestablishaframeworkforbuildingethicalAI.Figure 1showstheinitialframeworkforbuildingethicalAI. Table 2. AI ethics AI Human Society
EthicsofAI PrinciplesofdevelopingAItointeractwithotherAIsethically PrinciplesofdevelopingAItointeractwithhumanethically PrinciplesofdevelopingAItofunctionethicallyinsociety EthicalAI HowAIshouldinteractwithotherAIsethically? HowAIshouldinteractwithhumansethically? HowAIshouldoperateethicallyinsociety?
1.3. why Should we Build Ethical AI? Recently,CriminalsusedAI-basedvoicetechnologytoimpersonateachiefexecutive’svoiceand demandafraudulenttransferof$243,000(Stupp,2019).Thisisnotanisolatedincident.PINDROP reporteda350%riseinvoicefraudbetween2013and2017(Livni,2019).AIvoiceimpersonation beingusedforfraudisnottheonlyconcern.Deepfake,whichisanapproachtosuperimposeand synthesizeexistingimagesandvideosontosourceimagesorvideosusingMachineLearning(ML),is alsobecomingcommon.Withdeepfake,humanfacescouldbesuperimposedonpornographicvideo contentandpoliticalleaderscanbeportrayedinvideostoinciteviolenceandpanic.Deepfakemay alsobeusedintheelectioncycletoinfluenceandbiastheAmericanelectorate(Libby,2019).In2017, researchersfromtheUniversityofWashingtoncreatedasyntheticObama,usinganeuralnetwork AItomodeltheshapeofObama’smouth(BBCNews,2017).Althoughtherewasnosecuritythreat fromtheUniversityofWashingtonexperiment,thedemonstrationillustrateswhatispossiblewith AI-alteredvideos.Fakenewsisanotherconcern.Forexample,anAIfaketextgeneratorwasdeemed toodangeroustobereleasedbyitscreators,OpenAI,forfearofmisuse.Undoubtedly,advancedAI agentscouldputindividuals,companies,andsocietiesatincreasedrisk. Humanrights,suchasprivacy,freedomofassociation,freedomofspeech,righttowork,non-discrimination,andaccesstopublicservices,shouldalwaysbeputinthefirstplace.However,the growinguseofAIinthecriminaljusticesystemmayhaveadiscriminationconcern.Therecidivism risk-scoringsoftwareusedacrossthecriminaljusticesystemshowsincidentsofdiscrimination basedonrace,gender,andethnicity.Forinstance,somedefendantsarefalselylabeledashighrisk becauseoftheirethnicity. Therighttoprivacy,whichisessentialtohumandignity,canalsobeaffectedbyAI.Asbigdata technologydeveloped,thecollectionofdatainterfereswiththerightstoprivacyanddatasecurity. Forinstance,MLmodelscanaccuratelysynthesizedataandestimatepersonalcharacteristics,such asgender,age,maritalstatus,andoccupation,fromcellphonelocationdata.Anotherexampleis governmentsurveillance.IntheU.S.,halfofalladultsarealreadyinlawenforcementfacialrecognition databases(Telcher,2018),whichthreatenstoendanonymity.Rightstofreedomofexpression, assembly,andassociationmayaccordinglybeaffected.Lastbutnotleast,therighttoworkandan adequatestandardofliving(AccessNow,2018)wouldbeaffected.Automationhasresultedinjobloss andjobdisplacementincertainindustriesandtherapidadvancementinAIwouldacceleratethistrend.
2. REVIEw oF ETHICAL GUIDELINES, FRAMEwoRKS, AND PRINCIPLES
AdvancedAIwillsparkunprecedentedbusinessgains,butalongtheway,governmentandindustry leaderswillhavetograpplewithasmorgasbordofethicaldilemmassuchasdataprivacyissues, machinelearningbias,publicsafetyconcerns,aswellasjobreplacementandunemployment rateproblems.Toguidetheirstrategiesindeveloping,adopting,andembracingAItechnologies, organizationsshouldconsiderestablishingAIethicsframeworks/guidelines.Someinstitutionshave startedworkonthisissueandpublishedsomeguidelines.Table3showseightinstitutionsthatwork onAIethicalframeworksorprinciplesandtheirobjectives.Table4showsthecontentofthoseethical frameworksandprinciples.
InTable5,wesummarizedthefrequencyofeachfactorinthoseframeworksshowninTable 4.Wecanseethatdifferentframeworksmayincludethesameorsimilarfactors,butalsoinclude differentconsiderations.ThestudyofethicalissuesofAIisstillanewareaandmorediscussionis neededtofinallyestablishtheframeworkofbuildingethicalAI.Inthenextsection,wewilldiscuss eachethicalissueindetail.
3. REVIEw oF ETHICAL ISSUES IN AI
AI,atthepresentstage,isreferredtoasNarrowAIorWeakAI.WeakAIcandowellinanarrow andspecializeddomain.TheperformanceofnarrowAIdependsmuchonthetrainingdataand programming,whichiscloselyrelatedtobigdataandhumans.TheethicalissuesofNarrowAI, thus,involvehumanfactors. “Adifferentsetofethicalissuesariseswhenwecontemplatethepossibilitythatsomefuture AIsystemsmightbecandidatesforhavingthemoralstatus”(BostromandYudkowsky,2014,p.5).
Table 3. Institutions’ works on AI ethics and their objectives
Resource Objective FutureofLifeInstitute (2017) Thisreportemphasizes“donoharm”.ItrequiresthedevelopmentofAItobenefitsociety,fostertrustandcooperation,andavoidcompetitiveracing. International AssociationofPrivacy Professionals(IAPP, 2018) Theproposedframeworkexploresriskstoprivacy,fairness,transparency,equality,and manyotherissuesthatcanbeamplifiedbybigdataandartificialintelligence.They provideanoverviewofhoworganizationscanoperatedataethicsandhowtoreflectethical considerationsindecisionmaking. InstituteofElectrical andElectronics Engineers(IEEE, 2019) TheproposeddesignlaysoutpracticesforsettingupAIgovernancestructure,including pragmatictreatmentofdatamanagement,affectivecomputing,economics,legalaffairs,and otherareas.Onekeypriorityistoincreasehumanwell-beingasametricforAIprogress. Besides,theIEEEprinciplerequireseveryoneinvolvedinthedesignanddevelopmentofAIis educatedtoprioritizeethicalconsiderations. ThePublicVoice (2018) TheproposedguidelinesaimtoimprovethedesignanduseofAI,maximizethebenefits ofAI,protecthumanrights,andminimizerisksandthreatsassociatedwithAI.Theyclaim thattheguidelinesshouldbeincorporatedintoethicalstandards,adoptedinnationallawand internationalagreements,andbuiltintothedesignofsystems. European Commission’sHigh-LevelExpertGroup onAI(European Commission,2019) TheguidelinesaredesignedtoguidetheAIcommunityinthedevelopmentanduseof “trustworthyAI”(i.e.,AIthatislawful,ethical,androbust).Theguidelinesemphasizefour principles:respectforhumanautonomy,preventionofharm,fairness,andexplicability. AI4People(Floridiet al.,2018) ThisframeworkintroducesthecoreopportunitiesandrisksofAIforsociety;presenta synthesisoffiveethicalprinciplesthatshouldundergirditsdevelopmentandadoption;and offer20concreterecommendations—toassess,todevelop,toincentivize,andtosupport goodAI—whichinsomecasesmaybeundertakendirectlybynationalorsupranational policymakers. UnitedNations Educational, Scientific,and CulturalOrganization (UNESCO,2017) Theproposedethicalprincipleaimstoprovidedecision-makerswithcriteriathatextend beyondpurelyeconomicconsiderations. Australia’sEthics Framework(Dawsonet al.,2019) Thisethicsframeworkhighlightstheethicalissuesthatareemergingorlikelytoemergein AustraliafromAItechnologiesandoutlinestheinitialstepstowardmitigatingthem.Thegoal ofthisdocumentistoprovideapragmaticassessmentofkeyissuestohelpfosterethicalAI developmentinAustralia.
Theyadoptthedefinitionofmoralstatusthat“Xhasmoralstatus=becauseXcountsmorallyinits ownright,itispermissible/impermissibletodothingstoitforitsownsake.”Fromthisperspective, onceAIhasmoralstatus,weshouldtreatitnotasamachine/system,butanobjectthathasequal rightsashumans.Thetechnologicalsingularity,whentechnologicalgrowthbecomesuncontrollable andirreversible,ishypothesizedtocomeasAIadvances.Ifithappens,humancivilizationwould beaffected,androbotrightsandconsciousnessshouldbeconsidered.Buttheseissuesarebeyond theconsiderationofthispaper.Thefollowingdiscussionmainlyfocusesonethicalissuesrelated tonarrowAI. ResearchonethicalissuesofAIfallsintothreecategories:featuresofAIthatmaygiveriseto ethicalproblems(Timmermansetal.,2010),humanfactorsthatcauseethicalrisks(Larson,2017), andsocialimpactofethicalAIissues.
3.1. Features of AI Give Rise to Ethical Issues
3.1.1. Transparency
Machinelearningisabrillianttool,butitishardtoexplaintheinnerprocessingofmachine learning, which is usually called the “black box”. The “black box” makes the algorithms mysteriouseventoitscreators.Thislimitspeople’sabilitytounderstandthetechnology,leads tosignificantinformationasymmetriesamongAIexpertsandusers,andhindershumantrust inthetechnologyandAIagents.Trustiscrucialinallkindsofrelationshipsandaprerequisite reasonforacceptance(Siau&Wang,2018).
Table 4. Ethical frameworks and principles from eight institutions
Resource Ethical Framework/Principle
FutureofLifeInstitute(2017) Safety,FailureTransparency,JudicialTransparency,Responsibility,ValueAlignment,HumanValues,PersonalPrivacy,LibertyandPrivacy,Shared Benefit,SharedProsperity,HumanControl,Non-subversion,AIArmsRace InternationalAssociationofPrivacy
Professionals(IAPP,2018) Dataethics,Privacy,Bias,Accountability,Transparency,HumanRights
InstituteofElectricalandElectronics
Engineers(IEEE,2019) HumanRights,Well-being,DataAgency,Effectiveness,Transparency,Accountability,AwarenessofMisuse,Competence ThePublicVoice(2018) RighttoTransparency;RighttoHumanDetermination;Identification Obligation;FairnessObligation;AssessmentandAccountabilityObligation; Accuracy,Reliability,andValidityobligation;DataQualityObligation;Public SafetyObligation;CybersecurityObligation;ProhibitiononSecretProfiling; ProhibitiononUnitaryScoring;TerminationObligation. EuropeanCommission’sHigh-Level ExpertGrouponAI(European Commission,2019) HumanAgencyandOversight,TechnicalRobustnessandSafety,Privacyand DataGovernance,Transparency,Diversity,SocietalandEnvironmentalWell-being,Accountability AI4People(Floridietal.,2018) Beneficence:promotingwell-being,preservingdignity,sustainingtheplanet; Non-maleficence:privacy,security,monitoringAIadvancement/capability; Autonomy:thepowertodecide;Justice:promotingprosperity,preserving solidarity;Explicability:enablingtheotherprinciplesthroughintelligibility andaccountability UnitedNationsEducational, Scientific,andCulturalOrganization (UNESCO,2017) HumanDignity,ValueofAutonomy,ValueofPrivacy,“Donoharm” Principle,PrincipleofResponsibility,ValueofBeneficence,ValueofJustice Australia’sEthicsFramework (Dawsonetal.,2019) GeneratesNet-benefits,RegulatoryandLegalCompliance,Fairness, Contestability,DoNoHarm,PrivacyProtection,Transparencyand Explainability,Accountability
Further,becauseoftheblackboxthathumansarenotabletointerpret,AImayevolvewithout humanmonitoringandguidance.Forexample,in2017,FacebookshutdownanAIenginebecause theyfoundthattheAIhadcreateditsownuniquelanguageandhumanscouldnotunderstandthe language(Bradley2017).WhetherhumanscancontrolAIagentsisabigconcern.Humansprefer AIagentstoalwaysdoexactlywhatwewantthemtodo.Forinstance,ifaguestasksaself-driving taxitodrivetotheairportasfastaspossible,thetaximaynotfollowthetrafficrules,butreachthe airportatthefastestspeed.Thisisnotwhatthecustomerwantsbutwhatthecustomeraskedfor literally.However,consideringthisproblemfromanotherperspective,ifwetreatAIagentsethically, isitethicalthatwecontrolwhatactionstheytakeandhowtheymakedecisions?
3.1.2. Data Security and Privacy
ThedevelopmentofAIagentsreliesheavilyonthehugeamountofdata,includingpersonaldataand privatedata.Almostalloftheapplicationdomainsinwhichdeeplearningissuccessful,suchasApple SiriandGoogleHome,haveaccesstomountainsofdata.Withmoredatageneratedinsocietiesand businesses,thereisahigherchancetomisusethesedata.Forinstance,ahealthrecordalwayscontains sensitiveinformation,whichifnotadequatelyprotected,arogueinstitutioncouldgainaccesstothat informationandharmthepatientspersonallyandfinancially.Thus,datamustbemanagedproperly topreventmisuseandmalicioususe(Timmermansetal.,2010).Tokeepdatasafe,eachactionto thedatashouldbedetailedandrecorded.Boththedataperseandthetransactionrecordmaycause privacy-relatedrisks.Itis,therefore,importanttoconsiderwhatshouldberecordedandwhoshould takechargeoftherecordingaction,andwhocanhaveaccesstothedataandrecords.
Table 5. Summary of factors of ethical frameworks
Factors\Institutions FLI IAPP IEEE TPV EUCE AI4P UNESCO AEF Total
Responsibility/Accountability 1 1 1 1 1 1 1 1 8 Privacy 1 1 1 1 1 1 1 7 Transparency 1 1 1 1 1 1 6 HumanValues/DoNoHarm 1 1 1 1 1 1 6 HumanWell-Being/Beneficence 1 1 1 1 4 Safety 1 1 1 3 Liberty/Autonomy 1 1 1 3 HumanControl 1 1 1 3 Bias/Fairness 1 1 1 3 SharedBenefit 1 1 2 AIArmsRace 1 1 2 Justice 1 1 2 Prosperity 1 1 Effectiveness 1 1 Accuracy 1 1 Reliability 1 1 Diversity 1 1 HumanDignity 1 1 RegulatoryAndLegalCompliance 1 1
3.1.3. Autonomy, Intentionality, and Responsibility Whethertherobotsareregardedasmoralagentsaffecttheinteractions(Sullins,2011).Tobeseen asrealmoralagents,robotshavetomeetthreecriteria:autonomy,intentionality,andresponsibility (Sullins,2011).Autonomymeansthatthemachinesarenotunderthedirectcontrolofanyother agents.Intentionalitymeansthatmachines“actinawaythatismorallyharmfulorbeneficialand theactionsareseeminglydeliberateandcalculated.”Responsibilitymeansthemachinesfulfillsome socialrolethatcarrieswithitsomeassumedresponsibilities.Intheveryclassictrolleycase,theone whocontrolsthetrolleyistheethicalproducer(Allenetal.,2006).Tocontinuetorunonthecurrent trackandkillfiveworkersortoturntoanothertrackandkillaloneworkerisahard-ethicalchoice forhumans.WhatchoiceshouldorwouldAImake?WhoshouldberesponsiblefortheAI’schoice? Themilitaryrobotsthattakechargeofbombdisposalareethicalrecipients.Isitethicalthathumans decidethedestinyoftheserobots?Humanethicsandmoralitytodaymaynotbeseenasperfectby futurecivilizations(BostromandYudkowsky,2014).Onereasonisthathumanscannotsolveallthe recognizedethicalproblems.Theotherreasonisthathumanscannotrecognizealltheethicalproblems.
3.2. Human Factors Give Rise to Ethical Issues
HumanBias,suchasgenderbias(Larson,2017)andracebias(KoolenandCranenburgh,2017),may beinheritedbyAI.AIagentsareonlyasgoodasthedatahumanputintothem. AIagentsarebeingtrainedbyhumansandusingdatasetsmadebyhumans,existingbiasesmaybe learnedbyAIagentsandexhibitedinrealapplications.OncebiaseddataareusedbytheAIagent,the biaswillbecomeanongoingproblem.Forinstance,softwareusedtopredictfuturecriminalsshowed biasagainstacertainrace(Bossmann,2016).Thebiascomesfromthetrainingdatathatcontains humanbiases.Thus,figuringouthowtoprogramandtrainAIagentswithouthumanbiasesiscritical. 3.2.1. Accountability WhenanAIagentfailsatacertainassignedtask,whoshouldberesponsible.Thismayleadtowhat isreferredtoas“theproblemofmanyhands”(Timmermansetal.,2010).WhenusinganAIagent, anundesirableconsequencemaybecausedbytheprogrammingcodes,entereddata,improper operation,orotherfactors.Whoshouldbetheresponsibleentityfortheundesirableconsequence --theprogrammer,thedataowner,ortheend-users? 3.2.2. Ethical Standards “Theultimategoalofmachineethicsistocreateamachinethatitselffollowsanidealethicalprinciple orsetofprinciples”(AndersonandAnderson,2007p.15).Itistheoreticallyeasybutpracticallyhard toformulateethicalprinciplesforAIagents.Withoutcomprehensiveandunbiasedethicalstandards, howcanhumanstrainamachinetobeethical?Further,howcanwemakecertainthatintelligent machinesunderstandethicalstandardsinthesamewaythatwedo?(WangandSiau,2019a).For instance,ifweprogramrobotstoalwaysperformnoharm,weshouldfirstmakesurethattherobots understandwhatharmis.Thisresultsinanotherproblem--whatshouldbetheethicalstandardsfor harm?Aglobaloruniversallevelofethicsisneeded.Toputsuchethicsintomachines,itisnecessary toreducetheinformationasymmetriesbetweenAIprogrammersandcreatorsofethicalstandards. WhileattemptingtoformulateethicalstandardsforAIandintelligentmachines,researchersand practitionersshouldtrytobetterunderstandexistingethicalprinciplessothattheywillbeableto applytheethicalprinciplestoresearchactivitiesandhelptraindeveloperstobuildethicalAIs(Wang andSiau,2018).
3.2.3. Human Rights Laws
Withouttraininginhumanrightslaws,softwareengineersmaywritecodesthatviolateandbreach keyhumanrightswithoutevenknowingit.Itiscrucialtoteachhumanrightslawstosoftware
engineers.Ensuringprivacy-by-designisimportantandmorecost-efficientthanthealternatives. AbetterknowledgeofhumanrightslawscanhelpAIdesignersandengineerseliminateoratleast alleviatethediscriminationandinvasionofprivacyissuesinAI.
3.3. Social Impact of Ethical Issues
3.3.1. Automation and Job Replacement
ThedebateonwhetherAIageandIndustry4.0(Siau,Xi,andZou,2019;WangandSiau,2019b) willcreatemorejobsoreliminatesomejobsisstillheated.Storiesoffactoryworkersbeing replacedbyautomatedsystemsandrobotsabound.SomearguethatAIwillalsocreatemillions ofnewjobswithmanyofthesejobsthatarenon-existencetoday.Nevertheless,theconcern isstillthereaboutthefutureworkforcedisruptionsintheageofAI,suchasthecooperation betweenhumansandAIagents.ThelabormarketwillbedisruptedandtransformedwithAI. Whatisnotentirelyclearisthespeedandscopeofthechange.Theterm“uselessclass”hasbeen suggestedbyHarari(2016).UniversalBasicIncome(UBI)hasbeenpilotedinsomecountries andFreedomDividend,whichisauniversalbasicincomeforallAmericanadultswithnostrings attached,isthecampaignplatformfora2020U.S.PresidentialcandidateAndrewYang.The originalintentionoftechnologydevelopmentistoassisthumansandimprovehumanlives.If automationandAIcausehugejobreplacementandunemployment,shouldwekeeptherapid paceoftechnologydevelopment?Also,howcanweprotecthumanrightsandhumanwell-being whilekeepingupwiththerapidevolutionsandrevolutionsoftechnology? 3.3.2. Accessibility Accessibility,asanethicalprinciple,referstowhethersystems,products,andservicesareavailable, accessible,andsuitableforallpeople,includingtheelderly,thehandicapped,andthedisabled. Consideringthecomplexityofnewtechnologyandhigh-techproducts,aswellastheaging populationsinsomecountries,theaccessibilityofnewtechnologywilldirectlyaffecthumanwell-being.Technologydevelopmentshouldbenefithumans.Butifonlyaportionofpeoplebenefit,isit ethicalandfair?Considerationmustbegiventodevelopingsystems,products,andservicesthatare accessibletoall,andthebenefitsofadvancedtechnologyshouldbefairlydistributedtoall(Wang andSiau,2019b).
3.3.3. Democracy and Civil Rights
UnethicalAIresultsinthefragmentationoftruthandeventuallossoftrust,andlossofsocietal supportforAItechnology.Thelossofinformedandtrustingcommunitiesdentsthestrengthsof democracies.Asdemocraciessufferandstructuralbiasesamplified,thefreeexerciseofcivilrights nolongerremainsuniformlyavailabletoall.AIethicsneedstotakeintoconsiderationdemocracy andcivilrights. Figure2summarizestheaboveethicalissuesinAI.Solvingtheseissuesproperlywillhelpto establishaframeworkforbuildingethicalAI. 4. DISCUSSIoNS
4.1. From Framework to Practice
Figure2establishestheframeworkforAIethicslistingthefactorsthatneedtobeconsideredindefining theethicsofAIinordertobuildethicalAI.EventhoughdefiningtheethicsofAIismultifaceted andconvoluted,puttingtheethicsofAIintopracticetobuildethicalAIisnoeasyfeattoo.What shouldethicalAIlooklike?Inthesimplestform,wemaydefinethatethicalAIshoulddonoharm tohumans.But,whatisharm?Whatconstituteshumanrights?Manyquestionsneedtobeanswered beforewecandesignandbuildethicalAI.ethicalsensitivitytrainingisrequiredtomakegoodethical
decisions.Intheory,AIshouldbeabletorecognizeethicalissues.IfAIiscapableofmakingdecisions, howcanwedesignanddevelopanAIthatissensitivetoethicalissues?Unfortunately,itisnoteasy toimplementinpracticeandtorealize.Long-termandsustainedeffortsareneeded.Nonetheless, understandingandrealizingtheimportanceofdevelopingethicalAIandstartingtoworkonitstep bysteparepositivestepsforward. Manyinstitutions,suchasGoogle,IBM,Accenture,Microsoft,andAtomium-EISMD,have startedworkingonbuildingethicalprinciplestoguidethedevelopmentofAI.InNovember2018, theMonetaryAuthorityofSingapore(MAS),togetherwithMicrosoftandAmazonWebServices, launchedtheFEATprinciples(i.e.,fairness,ethics,accountability,andtransparency)fortheuseof AI.Academics,practitioners,andpolicymakersshouldworktogethertowidentheengagementto establishethicalprinciplesforAIdesign,development,anduse. Withtheframeworksandprinciples,protectiveguardrailstoensureethicalbehaviorsareneeded. Goodgovernanceisnecessarytoenforcetheimplementationandadherencetothoseethicalprinciples, andalegalvoidiswaitingtobefilledbyregulatoryauthorities(Hanna,2019).Eitherbasedoncase laworaccomplishedvialegislativeandregulatoryobligations,theselegalandregulatoryinstruments willbecriticaltothegoodgovernanceofAI,whichhelpstoimplementandenforceethicsofAIto enablethedevelopmentofethicalAI. Toprotectthepublic,theU.S.haslongenactedregulatoryinstruments,suchasrulesagainst discrimination,equalemploymentopportunity,HIPAATitleII,CommercialFacialRecognition PrivacyAct,andAlgorithmicAccountabilityAct.Alltheseinstrumentswouldbeusefulinguiding thedevelopmentoflegalandregulatorypoliciesandframeworksforAIethics. Inadditiontothelegalandgovernmentrules,self-regulationplaysanimportantroleaswell. Communicationandinformationdisclosurecanhelpsocietyasawholetoensurethedevelopment anddeploymentofethicalAI.Fosteringdiscussionforumsandpublishingethicalguidelinesby companies,industries,andpolicymakerscanhelptoeducateandtrainthepublicinunderstandingthe benefitsofAIanddispellingmythsandmisconceptionsaboutAI.Besides,havingabetterknowledge oflegalframeworksonhumanrights,strengtheningthesenseofsecurity,andunderstandingthe ethicalissuesrelatedtoAI,canfostertrustinAIandenablethedevelopmentofethicalAImore efficientlyandeffectively.
4.2. ways to Educate AI to Be Ethical
Moor(2006)indicatesthreepotentialwaystotransferAItoethicalagents:totrainAIinto“implicit ethicalagents”,“explicitethicalagents”,and“fullethicalagents”.Implicitethicalagentsmean
constrainingthemachine’sactionstoavoidunethicaloutcomes.Explicitethicalagentsmeanstating explicitlywhatactionisallowedandwhatisforbidden.Fullethicalagentsmeanmachines,ashumans, haveconsciousness,intentionality,andfreewill.Animplicitethicalagentcanrestrictthedevelopment ofAI.Anexplicitethicalagentiscurrentlygettingthemostattentionandisconsideredtobemore practical(AndersonandAnderson,2007).AfullethicalagentisstillanR&Dinitiativeandoneis notsurewhenafullethicalagentwillbeareality. Whenafullethicalagentisrealized,howtotreatanAIagentthathasconsciousness,moral sense,emotion,andfeelingswillbeanothercriticalconsideration.Forinstance,isitethicalto“kill” (shutdown)anAIagentifitreplaceshumanjobsorevenendangershumanlives?Isitethicalto deployrobotsintoadangerousenvironment?Thesequestionsareintertwinedwithhumanethics andmoralvalues.
4.3. Tradeoff Between AI Ethics and AI Advancement
President-electoftheEuropeanCommissionmadeclearinherrecentlyunveiledpolicyagendathat thecornerstoneoftheEuropeanAIplanwillbetoensurethat“AImadeinEurope”ismoreethical thanAImadeanywhereelseintheworld.TheEuropeanCommissionisnottheonlyonethatis concernedaboutAIethics.ManycountriesarealsoworkingonAIethics.U.S.agenciessuchas theDepartmentofDefenseandtheDepartmentofTransportationhavelaunchedtheirinitiatives toensuretheethicaluseofAIwithintheirrespectivedomains.InChina,thegovernment-backed BeijingAcademyofArtificialIntelligencehasdevelopedtheBeijingAIPrinciplesthatrivalthose ofothercountries,andtheChineseAssociationforArtificialIntelligencehasalsodevelopeditsown ethicsguidelines.Manynon-Europeancountries,includingtheUnitedStates,havesignedontothe OrganizationforEconomicCo-operationandDevelopment’s(OECD)AIPrinciplesfocusingon “responsiblestewardshipoftrustworthyAI.” However,themakersandresearchersofAIatthistimearelikelytopaymoreattentiontohard performancemetrics,suchassafetyandreliability,orsofterperformancemetrics,suchasusability andcustomersatisfaction.Morenebulousconceptslikeethicsarenotyetthemosturgentconsideration –especiallywiththeintensecompetitionbetweencompaniesandbetweennations. Further,whilesomeconsumersmaypaylipservicetoethicaldesign,theirwordsdonotmatch theiractions.Forexample,amongconsumerswhosaidtheydistrusttheInternet,only12%reportusing technologicaltools,suchasvirtualprivatenetwork,toprotecttheirdata,accordingtoaworldwideIpsos survey(CIGI-Ipsos,2019).Instead,themostimportantfactorsinfluencingconsumers’purchasing decisionsarestillthepriceandquality.Rightnow,consumerscaremoreaboutwhatAIcandorather thanwhetherallAI’sactionsareethical. ThissituationmayputcompaniesandinstitutionswhicharedevelopingAIinatradeoffsituation --whethertofocusonAIadvancementtorealizeprofitmaximization,ortofocusonAIethicsto ensuresocietalbenefitsfromAIinnovations. 5. CoNCLUSIoN UnderstandingandaddressingethicalandmoralissuesrelatedtoAIisstillintheinfancystage.AI ethicsisnotsimplyabout“rightorwrong”,“goodorbad”,and“virtueandvice”.Itisnotevena problemthatcanbesolvedbyasmallgroupofpeople.However,ethicalandmoralissuesrelatedto AIarecriticalandneedtobediscussednow.Thisresearchaimstocallattentiontotheurgentneed forvariousstakeholderstopayattentiontotheethicsandmoralityofAIagents.Whileattemptingto formulatetheethicsofAItoenablethedevelopmentofethicalAI,wewillalsounderstandhuman ethicsbetter,improvetheexistingethicalprinciples,andenhanceourinteractionswithAIagentsinthis AIage.AIethicsshouldbethecentralconsiderationindevelopingAIagentsandnotanafterthought. ThefutureofhumanitymaydependonthecorrectdevelopmentofAIethics!
REFERENCES
Allen,C.,Wallach,W.,&Smit,I.(2006).Whymachineethics?IEEE Intelligent Systems,21(4),12–17. doi:10.1109/MIS.2006.83
Anderson,M.,&Anderson,S.L.(2007).Machineethics:Creatinganethicalintelligentagent.AI Magazine, 28(4),15–26.
Anderson, M., & Anderson, S. L. (2011).Machine ethics. Cambridge University Press. doi:10.1017/ CBO9780511978036
Asimov,I.(1950).Runaround.InI,Robot(TheIsaacAsimovCollectionEd.).NewYorkCity:Doubleday. Blanchard,K.,&Peale,N.V.(2011).The power of ethical management.Randomhouse.
Bossmann,J.(2016).Top9ethicalissuesinartificialintelligence.WorldEconomicForum.Retrievedfrom https://www.weforum.org/ethical-issues-in-AI Bostrom,N.,&Yudkowsky,E.(2014).Theethicsofartificialintelligence.InTheCambridgehandbookof artificialintelligence(pp.316-334).CambridgePress.doi:10.1017/CBO9781139046855.020 Bradley,T.(2017).FacebookAICreatesItsOwnLanguageInCreepyPreviewofOurPotentialFuture.Forbes. Retrievedfromhttps://www.forbes.com/sites/tonybradley/2017/07/31/facebook-ai-creates-its-own-language-in-creepy-preview-of-our-potential-future/#45c65554292c Churchill,L.R.(1999).AreWeProfessionals?ACriticalLookattheSocialRoleofBioethicists.Daedalus, 253–274.PMID:11645877 CIGI-Ipsos.(2019).2019CIGI-IpsosGlobalSurveyonInternetSecurityandTrust.Retrievedfromhttp://www. cigionline.org/internet-survey-2019 Dawson,D.,Schleiger,E.,Horton,J.,McLaughlin,J.,Robinson,C.,Quezada,G.,...Hajkowicz,S.(2019). ArtificialIntelligence:Australia’sEthicsFramework.Data61CSIRO,Australia. EuropeanCommission.(2019).EthicsguidelinesfortrustworthyAI.Retrievedfromhttps://ec.europa.eu/digital-single-market/en/news/ethics-guidelines-trustworthy-ai Floridi,L.,Cowls,J.,Beltrametti,M.,Chatila,R.,Chazerand,P.,Dignum,V.,&Vayena,E.et al.(2018). AI4People—anethicalframeworkforagoodAIsociety:Opportunities,risks,principles,andrecommendations. Minds and Machines,28(4),689–707.doi:10.1007/s11023-018-9482-5PMID:30930541
FutureofLifeInstitute.(2017).AsilomarAIPrinciples.Retrievedfromhttps://futureoflife.org/ai-principles/?cn-reloaded=1
Hanna,M.(2019).Wedon’tneedmoreguidelinesorframeworksonethicalAIuse.It’stimeforregulatory action.BrinktheEdgeofRisk.Retrievedfromhttps://www.brinknews.com/we-dont-need-more-guidelines-or-frameworks-on-ethical-ai-use-its-time-for-regulatory-action/
Harari,Y.N.(2016).Homo Deus: a brief history of tomorrow.RandomHouse.
IAPP.(2018).WhitePaper--BuildingEthicsintoPrivacyFrameworksforBigDataandAI.Retrievedfrom https://iapp.org/resources/article/building-ethics-into-privacy-frameworks-for-big-data-and-ai/ IEEE.(2019).EthicallyalignedDesign.Retrievedfromhttps://ethicsinaction.ieee.org/ Koolen,C.,&vanCranenburgh,A.(2017).ThesearenottheStereotypesyouarelookingFor:BiasandFairnessin AuthorialGenderAttribution.ProceedingsoftheFirstACLWorkshoponEthicsinNaturalLanguageProcessing (pp.12-22).AcademicPress.doi:10.18653/v1/W17-1602 Larson,B.N.(2017).Genderasavariableinnatural-languageprocessing:Ethicalconsiderations. Libby,K.(2019).ThisBillHaderDeepfakevideoisAmazing.It’salsoTerrifyingforourFuture.Popular Mechanics.Retrievedfromhttps://www.popularmechanics.com/technology/security/a28691128/deepfake-technology/
Livni,E.(2019).AnewkindofcybercrimeusesAIandyourvoiceagainstyou.Quartz.Retrievedfromhttps:// qz.com/1699819/a-new-kind-of-cybercrime-uses-ai-and-your-voice-against-you/ MarkkulaCenterforAppliedEthics.(2015).Aframeworkforethicaldecisionmaking.SantaClaraUniversity. Retrievedfromhttps://www.scu.edu/ethics/ethics-resources/ethical-decision-making/a-framework-for-ethical-decision-making/ Moor,J.H.(1985).Whatiscomputerethics?Metaphilosophy,16(4),266–275.doi:10.1111/j.1467-9973.1985. tb00173.x
Moor,J.H.(2006).Thenature,importance,anddifficultyofmachineethics.IEEE Intelligent Systems,21(4), 18–21.doi:10.1109/MIS.2006.80 Nalini,B.(2019).TheHitchhiker’sGuidetoAIEthics.Medium.Retrievedfromhttps://towardsdatascience. com/ethics-of-ai-a-comprehensive-primer-1bfd039124b0 BBCNews.(2017).FakeObamacreatedusingAIvideotool[YouTubevideo].BBCNews.Retrievedfrom https://www.youtube.com/watch?v=AmUC4m6w1wo Now,A.(2018).Humanrightsintheageofartificialintelligence.Accessnow.org.Retrievedfromhttps://www. accessnow.org/cms/assets/uploads/2018/11/AI-and-Human-Rights.pdf
Paul,R.,&Elder,L.(2006).The Miniature Guide to Understanding the Foundations of Ethical Reasoning. UnitedStates:FoundationforCriticalThinkingFreePress.P.NP.
Resnik,D.B.(2011).Whatisethicsinresearchandwhyisitimportant.National Institute of Environmental Health Sciences,1(10),49–70.
Siau,K.,&Wang,W.(2018).BuildingTrustinArtificialIntelligence,MachineLearning,andRobotics.Cutter Business Technology Journal.,31(2),47–53.
Siau,K.,Xi,Y.,&Zou,C.(2019).Industry4.0:ChallengesandOpportunitiesinDifferentCountries.Cutter Business Technology Journal.,32(6),6–14.
Stupp,C.(2019).FraudstersUsedAItoMimicCEO’sVoiceinUnusualCybercrimeCase.TheWallStreet Journal. Retrieved from https://www.wsj.com/articles/fraudsters-use-ai-to-mimic-ceos-voice-in-unusual-cybercrime-case-11567157402
Sullins,J.P.(2011).Whenisarobotamoralagent.Machine ethics,151-160.
Telcher,J.G.(2018).Whatdofacialrecognitiontechnologiesmeanforourprivacy?The New York Times. Retrievedfromhttps://www.nytimes.com/2018/07/18/lens/what-do-facial-recognition-technologies-mean-for-our-privacy.html?nytapp=true&smid=nytcore-ios-share
ThePublicVoice.(2018).UniversalGuidelinesforArtificialIntelligence.Retrievedfromhttps://thepublicvoice. org/ai-universal-guidelines/
Timmermans,J.,Stahl,B.C.,Ikonen,V.,&Bozdag,E.(2010).Theethicsofcloudcomputing:Aconceptual review.Proceedings of theIEEE Second International Conference Cloud Computing Technology and Science (pp.614-620).IEEEPress.doi:10.1109/CloudCom.2010.59
UNESCO.(2017).ReportofWorldCommissionontheEthicsofScientificKnowledgeandTechnologyon RoboticsEthics.Retrievedfromhttps://unesdoc.unesco.org/ark:/48223/pf0000253952
Wang,W.,&Siau,K.(2018).Ethical and moral issue with AI – a case study on healthcare robots. AMCIS 2019 proceedings.AcademicPress.
Wang,W.,&Siau,K.(2019a).ArtificialIntelligence,MachineLearning,Automation,Robotics,FutureofWork andFutureofHumanity:AReviewandResearchAgenda.Journal of Database Management,30(1),61–79. Wang,W.,&Siau,K.(2019b).Industry4.0:EthicalandmoralPredicaments.Cutter Business Technology Journal.,32(6),36–45.
87
Keng Siau is Chair of the Department of Business and Information Technology at the Missouri University of Science and Technology. Previously, he was the Edwin J. Faulkner Chair Professor and Full Professor of Management at the University of Nebraska-Lincoln (UNL), where he was Director of the UNL-IBM Global Innovation Hub. Dr. Siau also served as VP of Education for the Association for Information Systems. He has written more than 300 academic publications, and is consistently ranked as one of the top information systems researchers in the world based on the h-index and productivity rate. Dr. Siau’s research has been funded by the U.S. National Science Foundation, IBM, and other IT organizations. He has received numerous teaching, research, service, and leadership awards, including from the International Federation for Information Processing Outstanding Service Award, the IBM Faculty Award, and the IBM Faculty Innovation Award. Dr. Siau received his Ph.D. in Business Administration from the University of British Columbia. He can be reached at siauk@mst.edu.
Weiyu Wang holds a Master of Science degree in Information Science and Technology and an MBA from the Missouri University of Science and Technology. Her research focuses on the impact of artificial intelligence (AI) on economy, society, and mental well-being. She is also interested in the governance, ethical, and trust issues related to AI. She can be reached at wwpmc@mst.edu.