自动驾驶汽车的安全性外文翻译中英文Word文档格式.docx

自动驾驶汽车的安全性外文翻译中英文Word文档格式.docx

《自动驾驶汽车的安全性外文翻译中英文Word文档格式.docx》由会员分享，可在线阅读，更多相关《自动驾驶汽车的安全性外文翻译中英文Word文档格式.docx（19页珍藏版）》请在冰点文库上搜索。

humandriver,”becauseitisnecessarilydefinedwithrespecttopopulation-leveldata.Atthelevelofindividualriskassessment,abodyofresearchhasshownthatmostdriversperceivethemselvestobesaferthantheaveragedriver（the 

better-than-averageeffect）. 

Method:

Usinganonlinesample,thisstudyexaminedU.S.drivers’ratingsoftheirownabilitytodrivesafelyandtheirdesiredlevelofsafetyforself-drivingvehicles. 

Results:

Thisstudyreplicatedthebetter-thanaverageeffectandshowedthatmostdriversstatedadesireforself-drivingcarsthataresaferthantheirownperceivedabilitytodrivesafelybeforetheywould:

（1）feelreasonablysaferidinginaself-drivingvehicle;

（2）buyaself-drivingvehicle,allotherthings（cost,etc.）beingequal;

and（3）allowself-drivingvehiclesonpublicroads. 

Conclusions:

Sincemostdriversbelievetheyarebetterthanaveragedrivers,thebenchmarkofachievingautomationthatissaferthanahumandriver（onaverage）maynotrepresentacceptablysafeperformanceofself-drivingcarsformostdrivers. 

Practicalapplications:

Ifperceivedlevelofsafetyisanimportantcontributortoacceptanceofself-drivingvehicles,thepopular“saferthanahumandriver”benchmarkmaynotbeadequateforwidespreadacceptance.

Keywords：

Human-automationinteraction，Trustinautomation，Self-drivingvehicles，Autonomousdriving，Vehicleautomation

Introduction

Advancesinengineeringandtechnologyhavemadetheprospectofmass-marketed,self-drivingvehiclesthesubjectofincreasingpublicdiscussion.Useracceptanceofautomationisanimportantdeterminantofwhetherthetechnologysucceedsorfails.Theextenttowhichthepublicandconsumerswillacceptself-drivingcarshasyettobedetermined,buttrust—“theattitudethatanagent[inthiscaseaself-drivingcar]willhelpachieveanindividual'

sgoalsinasituationcharacterizedbyuncertaintyandvulnerability”（Lee&

See,2004,p.51）—willbeanimportantfactor（Choi&

Ji,2015;

alsosee 

KaurandRampersad,2018, 

Kö

nigandNeumayr,2017）.Perceivedriskinfluencestrustinautomation,whichinturninfluencestheextenttowhichpeoplewillrelyonautomatedsystems（Hoff&

Bashir,2015）.Astudyofperceptionsofself-drivingcarsonTwitterfoundthatriskswerementionedthreetimesmoreoftenthanbenefitsinsocialmediadiscourse（Kohl,Mostafa,Bö

hm,&

Krcmar,2017）.Further,surveysconsistentlyhaveindicatedthatpeoplehaveskepticismandapprehensionsaboutvehicleautomationthatarerelatedtosafety（e.g., 

NielsenandHaustein,2018, 

Piaoetal.,2016, 

SchoettleandSivak,2014, 

Zmudetal.,2016）,witharecentsurvey（AmericanAutomobileAssociation,2018）reportingthatover70%ofU.S.driversfearridinginafullyautomatedvehicle.Clearly,safetywillbeanimportantfactorthatwillaffectacceptanceofself-drivingcars,butthecriterionforacceptablesafetyinvehicleautomationforthegeneralpubliciscurrentlyunknown.

Onepotentialsafetybenchmarkthathasemergedinthediscourseonvehicleautomationistocomparetheperformanceofself-drivingcarstohumandriverstoestablishwhethertheautomationissaferthanahumandriver.This 

saferthanahumandrivercriterion 

（SHDC）hasbecomeubiquitousinthediscoursesurroundingvehicleautomation.Forinstance,GoogleNews（i.e.,news-only）websearchesconductedonNovember17,2018forthephrases“saferthanahumandriver”and“saferthanhumandrivers”（bothsearchesinquotationmarks）returned711and1540respectiveresults—nearlyallofwhichreportednewsandanalysisonvehicleautomation.Thesereportsandcommentariescontainclaimsthatvehicleautomationalreadyhasmet（e.g.,byElonMusk,asreportedin 

McGoogan,2016）,hasnotmet（e.g., 

McArdle,2018）,and/oreventuallywillmeet（e.g., 

Hetherington,2018）theSHDC.Onesurvey（SchoettleandSivak,2014）showedthataround75%ofparticipantsweremoderatelytohighlyconcernedabout“self-drivingvehiclesnotdrivingaswellashumandriversingeneral,”whereasanother（Piaoetal.,2016）indicatedthatonly25%ofparticipantsbelievedthatautomatedvehicleswillbe“saferthanhumandrivenvehicles.”Indeed,muchofthediscussionofacceptablelevelsofsafetyforself-drivingcarsseemsto 

hinge 

uponwhethertheymeettheSHDC（e.g., 

Teoh&

Kidd,2017）,perhapsbecausethiscriterionpresentsastark,binarydelineationbetweenoutcomessuperiororinferiortothecurrentstatusquo.

TheSHDCdefinessafetyrelativetopopulation-leveldataparameters,whichmeansthecriterionisactually“saferthan 

humandriver.”Indeed, 

KalraandGroves（2017） 

estimatedthatwidespreaddeploymentofhighlyautomatedvehiclesthatareevensomewhat（i.e.,10%）saferthantheaveragehumandriverwouldsavemanylives（uptohalfamillion,dependingonvariousfactors）inthedecadesfollowingtheirintroduction.Assuch,thenormative,rationaldecisionforminimizinglossoflifeinthelong-termwouldbeforalldriverstousehighlyautomatedvehiclesassoonastheautomationhasbecomemarginallysaferthantheaveragehumandriver.

Atthelevelofindividualdecision-making,however,itisnotcertainthatvehicleautomationthatissaferthantheaveragedriverwillcompelwidespreadacceptance,adoption,andpurchaseofvehicleswithself-drivingcapabilities.Humandecision-makingdoesnotfollownormativeprobabilisticprocesses（forareview,see 

Beach&

Lipshitz,1993）.Perceivedriskmaybeinfluencedbyanumberof 

cognitivebiasesthatleadindividualstobelievetheirownpersonalriskisdifferentfromnormativeprobabilitiesbasedonpopulation-leveldata.Acrossawidevarietyofperformanceandtraitdomains,individualstendtoviewthemselvesasbetterthanaverage–aphenomenoncalled 

thebetter-than-averageeffect 

（forareview,see 

Alicke&

Govorun,2005）.

Inthecontextofdriving,researchconsistentlyhasshownthatmostdriversbelievethemselvestobeasaferdriverthanboththeaveragehumandriverandthedriversintheirownpeergroups（FinnandBragg,1986, 

Harré

etal.,2005, 

MarottoliandRichardson,1998, 

McCormicketal.,1986）. 

Svenson（1981） 

showedthat88%ofasampleofU.S.driversand77%ofasampleofSwedishdriversratedthemselvesassaferthanthemediandriver. 

McKenna,Stanier,andLewis（1991） 

showedthatparticipantsratedthemselvesasmoreskillfulthantheaveragedriveracrossanumberofdrivingtasksthataffectsafety（e.g.,payingattentiontoothervehicles,judgingdistancesrequiredtostop,andmindingappropriatespeedsfordrivingconditions）.InasampleofUKdrivers, 

Horswill,Waylen,andTofield（2004） 

showedthatparticipantsratedthemselvesasmoresafe,moreskillful,andlessaccidentpronethanboththeirdemographicallymatchedpeersandtheaveragedriver.

McKenna（1993） 

askedpeopletoratetheirlikelihoodofbeinginvolvedinanaccidentwhileridinginavehicleasapassengerversuswhiledrivingthevehicle.Peoplebelievedthattheyhadanaveragechanceofbeinginvolvedinanaccidentasapassenger,buttheyreportedthattheyhadalessthanaveragechanceofbeinginvolvedinanaccidentasthedriverofacar.Inotherwords,peopleseemedtobelievethatpopulation-levelstatisticsaboutaccidentsappliedtootherdrivers,butnottothemselves.McKennaconcludedthatthemechanismforthisbiasistheillusionofcontrol（ratherthanalternativeexplanationssuchasunrealisticoptimism）.Relatedfindingsreportedby 

DeJoy（1989） 

suggestedthatunrealisticoptimismstemsfromthetendencyforpeopletooverestimatetheextenttowhichtheyhavecontrolovercircumstancethatcouldleadtoanaccidentwhiledriving.Correspondingly,riskdenial（believingone'

sownriskislessthanthatofthegeneralpopulation）increasesasperceivedcontroloverriskincreases（Sjö

berg,2000）.

Abiastowardinflatedperceptionsofsafetywhileoneisincontrolofavehiclecouldbeproblematicforacceptanceofself-drivingcars.Researchhassuggestedthatpeoplehaveastrongdesiretobeabletotakecontrolbackfromautomatedsystems（Kö

nigandNeumayr,2017, 

Nees,2016, 

SchoettleandSivak,2016）,likelybecausetheyfeelsaferwhentheyareincontrolofthevehicle（e.g., 

DeJoy,1989, 

Sjö

berg,2000）ratherthanridingasapassivepassenger（see（McKenna,1993）.Althoughthenormative,rationaldecisionatapopulationlevelmaybeforalldriverstorelinquishcontroltoself-drivingcarsassoonasautomationissaferthantheaveragehumandriver（see 

Kalra&

Groves,2017）,the 

realization 

ofpopulation-levelbenefits（e.g.,reducedtrafficfatalities）willrequirethatindividualsacceptanduseorbuyself-drivingcars.Ifmostindividualsbelievetheyareabove-averagedriverswithrespecttosafetyduetobiasesrelatedtotheillusionofcontroloverrisks,thentheymayrequireself-drivingcarstomeetahigherbenchmarkthantheSHDCbeforetheywillacceptthisnewtechnology.Specifically,itseemsthatpeoplewillwantself-drivingcarstobesaferthantheybelievethemselvestobe.Forexample,arecentstudy（Liu,Yang,&

Xu,2019）modeledriskacceptabilityforself-drivingvehiclesinasampleofresponsesfromChineseparticipants.Theirresultssuggestedthattheminimumacceptablesafetycriterionmaybe4–5timessaferthanhumandrivers.

Thecurrentstud