《世卫组织-人工智能健康伦理与治理：大型多模式模型指南（英）-2024.docx》由会员分享，可在线阅读，更多相关《世卫组织-人工智能健康伦理与治理：大型多模式模型指南（英）-2024.docx（90页珍藏版）》请在三一办公上搜索。
1、IanizationEthicsandgovernanceofartificialintelligenceforhealthGuidanceonlargemulti-modalmodelsWorldHealthOiganizationEthicsandgovernanceofartificialintelligenceforhealthGuidanceonlargemulti-modalmodelsISBN978-92-4-008475-9(electronicversion)ISBN978-92-4-008476-6(printversion)WorldHealthOrganization2
2、024Somerightsreserved.ThisworkisavailableundertheCreativeCommonsAttribution-NonCommerciaI-ShareAIike3.0IGOlicence(CCBY-NC-SA3.0IGO;https:/creativecommons.Org/licenses/by-nc-sa/3.0/igo).Underthetermsofthislicence,youmaycopy,redistributeandadapttheworkfornon-commercialpurposes,providedtheworkisappropr
3、iatelycited,asindicatedbelow.Inanyuseofthiswork,thereshouldbenosuggestionthatWHOendorsesanyspecificorganization,productsorservices.TheuseoftheWHOlogoisnotpermitted.Ifyouadaptthework,thenyoumustlicenseyourworkunderthesameorequivalentCreativeCommonslicence.Ifyoucreateatranslationofthiswork,youshouldad
4、dthefollowingdisclaimeralongwiththesuggestedcitation:ThistranslationwasnotcreatedbytheWorldHealthOrganization(WHO).WHOisnotresponsibleforthecontentoraccuracyofthistranslation.TheoriginalEnglisheditionshallbethebindingandauthenticedition,Anymediationrelatingtodisputesarisingunderthelicenceshallbecond
5、uctedinaccordancewiththemediationrulesoftheWorldIntellectualPropertyOrganization,(http:/www.wipo.int/amc/en/mediation/rules/).Suggestedcitation.Ethicsandgovernanceofartificialintelligenceforhealth.Guidanceonlargemulti-modalmodels.Geneva:WorldHealthOrganization;2024.1.icence:CCBY-NC-SA3.0IGO.Catalogu
6、ing-In-Publication(CIP)data.CIPdataareavailableathttp:/apps.who.int/iris.Sales,rightsandlicensing.TopurchaseWHOpublications,seehttp:/apps.who.int/bookorders.TosubmitrequestsforCommercicensing,seehttp:/www.who.int/about/licensing.Third-partymaterials.Ifyouwishtoreusematerialfromthisworkthatisattribut
7、edtoathirdparty,suchastables,figuresorimages,itisyourresponsibilitytodeterminewhetherpermissionisneededforthatreuseandtoobtainpermissionfromthecopyrightholder.Theriskofclaimsresultingfrominfringementofanythird-partyownedcomponentintheworkrestssolelywiththeuser.Generaldisclaimers.Thedesignationsemplo
8、yedandthepresentationofthematerialinthispublicationdonotimplytheexpressionofanyopinionwhatsoeveronthepartofWHOconcerningthelegalstatusofanycountry,territory,cityorareaorofitsauthorities,orconcerningthedelimitationofitsfrontiersorboundaries.Dottedanddashedlinesonmapsrepresentapproximateborderlinesfor
9、whichtheremaynotyetbefullagreement.Thementionofspecificcompaniesorofcertainmanufacturers*productsdoesnotimplythattheyareendorsedorrecommendedbyWHOinpreferencetoothersofasimilarnaturethatarenotmentioned.Errorsandomissionsexcepted,thenamesofproprietaryproductsaredistinguishedbyinitialcapitalletters.Al
10、lreasonableprecautionshavebeentakenbyWHOtoverifytheinformationcontainedinthispublication.However,thepublishedmaterialisbeingdistributedwithoutwarrantyofanykind,eitherexpressedorimplied.Theresponsibilityfortheinterpretationanduseofthemateriallieswiththereader.InnoeventshallWHObeliablefordamagesarisin
11、gfromitsuse.Graphicsdesign:JoannaSleigh(ETHZurich,Zurich,Switzerland)1.ayout:ImprimerieCentrale(1.uxembourg)ContentsAcknowledgementsvAbbreviationsviiExecutivesummaryviii1 Introduction11.1 Significanceof1.MMs31.2 WHOguidanceonethicsandgovernanceofAlforhealth4,Applications,challengesandrisksof1.MMs72
12、Applicationsandchallengesofuseof1.MMsinhealth82.1 Diagnosisandclinicalcare82.2 Patient-centredapplications122.3 Clericalfunctionsandadministrativetasks162.4 Medicalandnursingeducation172.5 Scientificandmedicalresearchanddrugdevelopment173 Riskstohealthsystemsandsocietyandethicalconcernsaboutuseof1.M
13、Ms203.1 Healthsystems203.2 Compliancewithregulatoryandlegalrequirements233.3 Societalconcernsandrisks24AEthicsandgovernanceof1.MMsinhealthcareandmedicine314 Designanddevelopmentofgeneral-purposefoundationmodels(1.MMs)344.1 Riskstobeaddressedduringthedevelopmentofgeneral-purposefoundationmodels(1.MMs
14、)344.2 Measuresdeveloperscantaketoaddressriskswithgeneral-purposefoundationmodels(1.MMs)354.3 Governmentlaws,policiesandpublicsectorinvestments394.4 Open-source1.MMs415 Provisionwithgeneral-purposefoundationmodels(1.MMs)455.1 Riskstobeaddressedwhenprovidingahealth-careserviceorapplicationwithagenera
15、l-purposefoundationmodel(1.MM)455.2 Measuresthatgovernmentscanintroducetoaddresssuchrisksandethicalprinciplesthatshouldbeupheld466 Deploymentwithgeneral-purposefoundationmodels(1.MMs)536.1 Riskstobeaddressedwhendeployingahealth-careserviceorapplicationwithageneral-purposefoundationmodel(1.MM)536.2 O
16、n-goingresponsibilitiesofdevelopersandprovidersduringdeployment546.3 Responsibilitiesofdeployers546.4 Governmentprogrammesandpractices557 1.iabilityfor1.MMs588 Internationalgovernanceof1.MMs60References62Annex.Methods77AcknowledgementsDevelopmentofthisWorldHealthOrganization(WHO)guidancewasledbyAndr
17、easReis(co-leadoftheHealthEthicsandGovernanceunitintheDepartmentofResearchforHealth)andSameerPujari(DepartmentofDigitalHealthandInnovation),undertheoverallguidanceofJohnReeder(Director,ResearchforHealth),Alain1.abrique(Director,DigitalHealthandInnovation)andJeremyFarrar(ChiefScientist).RohitMalpani(
18、consultant,France)wastheleadwriter.Theco-chairsoftheWHOexpertgrouponethicsandgovernanceofAlforhealth,EffyVayena(ETHZurich,Switzerland)andParthaMajumder(IndianStatisticalInstituteandNationalInstituteofBiomedicalGenomics,India),providedoverallguidanceondraftingofthereportandleadershipoftheexpertgroup.
19、WHOisgratefultothefollowingindividualswhocontributedtodevelopmentofthisguidance.WHOexpertgrouponethicsandgovernanceofAlforhealthNajeebAlShorbaji,eHealthDevelopmentAssociation,Amman,Jordan;MariaPazCanales,GlobalPartnersDigital,SantiagodeChile,Chile;ArisaEma,UniversityofTokyo,Tokyo,Japan;AmelGhouila,B
20、ill&MelindaGatesFoundation,Seattle(WA),USA;JenniferGibson,WHOCollaboratingCentreforBioethics,UniversityofToronto,Toronto,Canada;KennethGoodman,InstituteofBioethicsandHealthPolicy,UniversityofMiamiMillerSchoolofMedicines,Miami(F1.)zUSA;MalavikaJayaram,DigitalAsiaHubzSingapore;DaudiJjingo,MakerereUniv
21、ersity,Kampala,Uganda;TzeYun1.eong,NationalUniversityOfSingaporezSingapore;AIexJohn1.ondon,CarnegieMellonUniversity,Pittsburgh(PA)zUSA;ParthaMajumder,IndianStatisticalInstituteandNationalInstituteofBiomedicalGenomics,Kolkata,India;ThilidziMarwalazUniversityOfJohannesburg,Johannesburg,SouthAfrica;Rol
22、iMathur,IndianCouncilofMedicalResearch,Bangalore,India;TimoMinssen,CentreforAdvancedStudiesinBiomedicalInnovation1.aw,Facultyof1.aw,UniversityofCopenhagen,Copenhagen,Denmark;AndrewMorris,HealthDataResearchUKz1.ondon,UnitedKingdom;DanielaPaolotti,ISIFoundation,Turin,Italy;JeromeSingh,UniversityofKwa-
23、ZuluNatal,Durban,SouthAfrica;JeroenvandenHoven,UniversityofDelft,Delft,Netherlands(Kingdomofthe);EffyVayenazETHZurich,Zurich,Switzerland;RobynWhittaker,UniversityofAuckland,Auckland,NewZealand;andYiZeng,ChineseAcademyofSciences,Beijing,China.ObserversDavidGruson,1.uminess,Pans,France;1.eeHibbard,Cou
24、ncilofEurope,Strasbourg,FranceExternalreviewersOrenAsman,TelAvivUniversity,TelAviv,Israel;I.GlennCohen,Harvard1.awSchool,Boston(MA),USA;AlexandrinePirlotdeCorbion,PrivacyInternational,1.ondon,UnitedKingdom;Rodrigo1.ins,FederalUniversityofPernambuco,Recife,Brazil;DougMcNairzDeputyDirector,IntegratedD
25、evelopment,Bill&MelindaGatesFoundation,Seattle(WA)zUSA;KeymanthriMoodleyzStellenboschUniversity,CapeTown,SouthAfrica;AmirTaI,TelAvivUniversity,TelAviv,Israel;TomWestzPrivacyInternational,1.ondon,UnitedKingdom.ExternalcontributorsBox2(Ethicalconsiderationsfortheuseof1.MMsbychildren)oftheguidancewasdr
26、aftedbyVijaythaMuralidharan,AlyssaBurgart,RoxanaDaneshjouandSherriRose,StanfordUniversity,Stanford(CA),USA.Box3(Ethicalconsiderationsassociatedwith1.MMsandtheirimpactonindividualswithdisabilities)oftheguidancewasdraftedbyYonahWelker,independentconsultant,Geneva,Switzerland.Allexternalreviewers,exper
27、tsandcontributorsdeclaredtheirinterestsinlinewithWHOpolicies.Noneoftheinterestsdeclaredwereassessedtobesignificant.WHOShadaAl-Salamah,TechnicalOfficer,DepartmentofDigitalHealthandInnovation,Geneva;MariamOtmaniDelBarrio,Scientist,SpecialProgrammeonTropicalDiseasesResearch,Geneva;MarceloDAgostino,Unit
28、Chief,InformationSystemsandDigitalHealth,WHORegionalOfficefortheAmericas,Washington(DC);JeremyFarrar,ChiefScientist,Geneva;ClaytonHamilton,TechnicalOfficer,WHORegionalOfficeforEurope,Copenhagen,Denmark;KanikaKalra,Consultant,DepartmentofDigitalHealthandInnovation,Geneva;AhmedMohamedAminMandilzCoordi
29、nator,ResearchandInnovation,WHORegionalOfficefortheEasternMediterranean,Cairo;IssaT.Matta,1.egalAffairs,Geneva;JoseEduardoDiazMendoza,Consultant,DepartmentofDigitalHealthandInnovation,Geneva;MohammedHassanNour,TechnicalOfficer,DepartmentofDigitalHealthandInnovation,WHORegionalOfficefortheEasternMedi
30、terranean,Cairo;DeniseSchaletzTechnicalOfficer,DepartmentofDigitalHealthandInnovation,Geneva;YuZhao,TechnicalOfficer,DepartmentofDigitalHealthandInnovation,Geneva.AbbreviationsAlartificialintelligence1.MMlargemulti-modalmodelUSAUnitedStatesofAmericaExecutivesummaryArtificialIntelligence(AI)referstot
31、hecapabilityofalgorithmsintegratedintosystemsandtoolstolearnfromdatasothattheycanperformautomatedtaskswithoutexplicitprogrammingofeverystepbyahuman.GenerativeAlisacategoryofAltechniquesinwhichalgorithmsaretrainedondatasetsthatcanbeusedtogeneratenewcontent,suchastext,imagesorvideo.Thisguidanceaddress
32、esonetypeofgenerativeAl,largemulti-modalmodels(1.MMs),whichcanacceptoneormoretypeofdatainputandgeneratediverseoutputsthatarenotlimitedtothetypeofdatafedintothealgorithm.Ithasbeenpredictedthat1.MMswillhavewideuseandapplicationinhealthcare,scientificresearch,publichealthanddrugdevelopment.1.MMsarealso
33、knownas/general-purposefoundationmodels”,althoughitisnotyetprovenwhether1.MMscanaccomplishawiderangeoftasksandpurposes.1.MMshavebeenadoptedfasterthananyconsumerapplicationinhistory.Theyarecompellingbecausetheyfacilitatehuman-computerinteractiontomimichumancommunicationandtogenerateresponsestoqueries
34、ordatainputsthatmayappearhuman-likeandauthoritative.Withrapidconsumeradoptionanduptakeandinviewofitspotentialtodisruptcoresocialservicesandeconomicsectors,manylargetechnologycompanies,start-upsandgovernmentsareinvestinginandCompetingtoguidethedevelopmentOfgenerativeAI.In2021,WHOpublishedcomprehensiv
35、eguidance(1)ontheethicsandgovernanceofAlforhealth.WHOconsulted20leadingexpertsinAl,whoidentifiedbothpotentialbenefitsandpotentialrisksofuseofAlinhealthcareandissuedsixprinciplesarrivedatbyconsensusforconsiderationinthepoliciesandpracticesOfgovernmentszdevelopers,andprovidersthatareusingAl.Theprincip
36、lesshouldguidethedevelopmentanddeploymentofAlinhealthcarebyawiderangeofstakeholders,includinggovernments,publicsectoragencies,researchers,companiesandimplementers.Theprinciplesare:(1)protectautonomy;(2)promotehumanwell-being,humansafetyandthepublicinterest;(3)ensuretransparency,zzexplainabilitywandi
37、ntelligibility;(4)fosterresponsibilityandaccountability;(5)ensureinclusivenessandequity;and(6)promoteAlthatisresponsiveandsustainable(Figure1).WHOisissuingthisguidancetoassistMemberStatesinmappingthebenefitsandchallengesassociatedwithuseof1.MMsforhealthandindevelopingpoliciesandpracticesforappropria
38、tedevelopment,provisionanduse.Theguidanceincludesrecommendationsforgovernance,withincompanies,bygovernmentsandthroughinternationalcollaboration,alignedwiththeguidingprinciples.Theprinciplesandrecommendations,whichaccountfortheuniquewaysinwhichhumanscanusegenerativeAlforhealth,arethebasisofthisguidan
39、ce.Figure 1: WHOconsensusethicalprinciplesforuseofAlforhealthProtectautonomy。2。FosterresponsibilityandaccountabilityPromotehumanwell-being,humansafetyandthepublicinterest国EnsureinclusivenessandequityEnsuretransparency,explainabilityandintelligibility魄PromoteAIthatisresponsiveandsustainable浸Applicati
40、ons,challengesandrisksoflargemulti-modalmodelsThepotentialapplicationsof1.MMsinhealthcarearesimilartothoseofotherformsofAl,yethow1.MMsareaccessedandusedisnew,withbothnovelbenefitsandrisksthatsocieties,healthsystemsandend-usersmaynotyetbepreparedtoaddressfully.Table1summarizesthemainapplicationsof1.M
41、Msandtheirpotentialbenefitsandrisks.Thesystemicrisksassociatedwithuseof1.MMsincluderisksthatcouldaffecthealthsystems(Table2).Broaderregulatoryandsystemicriskscouldemergewithuseof1.MMs.Oneconcern(beingexaminedbyseveraldataprotectionauthorities)iswhether1.MMscomplywithexistinglegalorregulatoryregimes,
42、includinginternationalhumanrightsobligations,andwithnationaldataprotectionregulations.Algorithmsmightnotcomplywithsuchlawsbecauseofthewayinwhichdataarecollectedtotrain1.MMs,themanagementandprocessingofdatathathavebeencollected(orputinto1.MMsbyendusers),thetransparencyandaccountabilityofentitiesthatd
43、evelop1.MMs,andthepossibilitythat1.MMszzhallucinate,1.MMscouldalsobenon-Compliantwithconsumerprotectionlaws.Broadersocietalrisksassociatedwiththegrowinguseof1.MMs(includingandbeyondtheuseofsuchalgorithmsinhealthcare)includethefactthat1.MMsareoftendevelopedanddeployedbylargetechnologycompanies,duepar
44、tlytothesignificantcomputing,data,humanandfinancialresourcerequiredfordevelopmentof1.MMs.Thismayreinforcethedominanceofthesecompaniesvis-a-vissmallerenterprisesandgovernmentswithrespecttothedevelopmentanduseofAlzincludingthefocusofAlresearchinthepublicandprivatesectors.Additionalconcernsaboutthepote
45、ntialdominanceoflargetechnologycompaniesincludeinsufficientcorporatecommitmenttoethicsandtransparency.NewvoluntaryTable1.Potentialbenefitsandrisksinvarioususesof1.MMSinhealthcareUsePotentialorproposedbenefitsPotentialrisksIDiagnosisandclinicalcareAssistinmanagingcomplexcasesandreviewofroutinediagnos
46、esReducethecommunicationworkloadofhealth-careproviders(keyboardliberation,ProvidenovelinsightsandreportsfromvariousunstructuredformsofhealthdataInaccurate,incompleteorfalseresponsesPoorqualitytrainingdataBiasoftrainingdataandresponses)AutomationbiasDegradationofskills(ofhealth-careprofessionals)Info
47、rmedconsent(ofpatients)Patient-guideduseGenerateinformationtoimproveunderstandingofamedicalcondition(asapatientorasacaregiver)VirtualhealthassistantClinicaltrialenrolmentInaccurate,incompleteorfalsestatementsManipulationPrivacy1.essinteractionbetweencliniciansandpatientsEpistemicinjusticeRiskofdeliv
48、eryofcareoutsidethehealthsystemClericalandadministrativetasksAssistwithpaperworkanddocumentationrequiredforclinicalcareAssistinlanguagetranslationCompletionofelectronichealthrecordsDraftclinicalnotesafterapatientvisitInaccuraciesanderrorsInconsistentresponsesdependingonpromptsMedicalandnursingeducationDynamictextssuitedtoeachstudentsneedsSimulatedconversationtoimprovecommunicationandtopractiseindiversesituationsandwithdiversepatientsResponsestoquestionsaccompaniedbychain-of-thoughtreasoningContributetoautomationbiasErrorsorfalseinformationunderm
