原创力文档- 1、本文档共34页，可阅读全部内容。
- 2、原创力文档（book118）网站文档一经付费（服务费），不意味着购买了该文档的版权，仅供个人/单位学习、研究之用，不得用于商业用途，未经授权，严禁复制、发行、汇编、翻译或者网络传播等，侵权必究。
- 3、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。如您付费，意味着您自己接受本站规则且自行承担风险，本站不退款、不进行额外附加服务；查看《如何避免下载的几个坑》。如果您已付费下载过本站文档，您可以点击 这里二次下载。
- 4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等，请点击“版权申诉”（推荐），也可以打举报电话：400-050-0827(电话支持时间：9:00-18:30)。
IssueBrief
AISafetyand
AutomationBias
TheDownsideof
Human-in-the-Loop
Authors
LaurenKahn
EmeliaS.Probasco
RonnieKinoshita
November2024
ExecutiveSummary
Automationbiasisthetendencyforanindividualtoover-relyonanautomated
system.Itcanleadtoincreasedriskofaccidents,errors,andotheradverseoutcomes
whenindividualsandorganizationsfavortheoutputorsuggestionofthesystem,even
inthefaceofcontradictoryinformation.
Automationbiascanendangerthesuccessfuluseofartificialintelligencebyeroding
theuser’sabilitytomeaningfullycontrolanAIsystem.AsAIsystemshave
proliferated,sotoohaveincidentswherethesesystemshavefailedorerredinvarious
ways,andhumanusershavefailedtocorrectorrecognizethesebehaviors.
Thisstudyprovidesathree-tieredframeworktounderstandautomationbiasby
examiningtheroleofusers,technicaldesign,andorganizationsininfluencing
automationbias.Itpresentscasestudiesoneachofthesefactors,thenofferslessons
learnedandcorrespondingrecommendations.
UserBias:TeslaCaseStudy
Factorsinfluencingbias:
●User’spersonalknowledge,experience,andfamiliaritywithatechnology.
●User’sdegreeoftrustandconfidenceinthemselvesandthesystem.
Lessonslearnedfromcasestudy:
●Disparitiesbetweenuserperceptionsandsystemcapabilitiescontributeto
biasandmayleadtoharm.
Recommendation:
●Createandmaintainqualificationstandardsforuserunderstanding.
Usermisunderstandingofasystem’scapabilitiesorlimitationsisa
significantcontributortoincidentsofharm.Sinceuserunderstandingis
criticaltosafeoperation,systemdevelopersandvendorsmustinvestin
clearcommunicationsabouttheirsystems.
CenterforSecurityandEmergingTechnology|
文档评论（0）