JESK. Technology Acceptance Modeling based on User Experience for Autonomous Vehicles. Yujun Cho 1, Jaekyu Park 1, Sungjun Park 2, Eui S.

Transcription

1 JESK J Ergon Soc Kore 2017; 36(2): eissn: Technology Acceptnce Modeling bsed on User Experience for Autonomous Vehicles Yujun Cho 1, Jekyu Prk 1, Sungjun Prk 2, Eui S. Jung 1 1 Kore University, Deprtment of Industril Mngement Engineering, Seoul, Nmseoul University, Deprtment of Industril Mngement Engineering, Cheonn, Corresponding Author Eui S. Jung Kore University, Deprtment of Industril Mngement Engineering, Seoul, Phone : Emil : ejung@kore.c.kr Received : October 07, 2016 Accepted : November 28, 2016 Objective: The purpose of this study ws to precede the cceptnce study bsed on utomtion steps nd user experience tht ws lcked in the pst study on the core technology of utonomous vehicle, ADAS. The first objective ws to construct the cceptnce model of ADAS technology tht is the core technology, nd drw fctors tht ffect behviorl intention through user experience-bsed evlution by pplying driving simultor. The second one ws to see the chnge of fctors on utomtion step of utonomous vehicle through the UX/UA score. Bckground: The number of vehicles with the introduction of ADAS is incresing, nd it cused chnge of interction between vehicle nd driver s utomtion is being developed on the prticulr drive fctor. For this reson, it is becoming importnt to study the technology cceptnce on how driver cn ctively ccept giving up some prts of utomted drive opertion nd hnding over the uthority to vehicle. Method: We orgnized the study model nd items through literture investigtion nd the scenrio ccording to the 4 stges of utomtion of utonomous vehicle, nd preceded cceptnce ssessment using driving simultor. Totl 68 men nd womn were prticipted in this experiment. Results: We drew results of Performnce Expectncy (PE), Socil Influence (SI), Perceived Sfety (PS), Anxiety (AX), Trust (T) nd Affective Stisfction (AS) s the fctors tht ffect Behviorl Intention (BI). Also the drwn fctors shows tht UX/UA score hs significnt difference sttisticlly ccording to the utomtion steps of utonomous vehicle, nd UX/UA tends to move up until the stge 2 of utomtion, nd t stge 3 it goes down to the lowest level, nd it increses little or stys stedy t stge 4. Copyright@2017 by Ergonomics Society of Kore. All right reserved. cc This is n open-ccess rticle distributed under the terms of the Cretive Commons Attribution Non-Commercil License ( cretivecommons.org/licenses/by-nc/3.0/), which permits unrestricted non-commercil use, distribution, nd reproduction in ny medium, provided the originl work is properly cited. Conclusion nd Appliction: First, we presented the cceptnce model of ADAS tht is the core technology of utonomous vehicle, nd it could be the bsis of the future cceptnce study of the ADAS technology s it verifies through user experience-bsed ssessment using driving simultor. Second, it could be helpful to the pproprite ADAS development in the future s drwing the chnge of fctors nd predicting the cceptnce level ccording to the utomtion stges of utonomous vehicle through UX/UA score, nd it could lso grsp nd void the problem tht ffect the cceptnce level. It is possible to use these study results s tools to test vlidity of function before ADAS offering compny lunches the products. Also it will help to prevent the problems tht could be cused when pplying the utonomous vehicle technology, nd to estblish technology tht is esily cceptble for drivers, so it will improve sfety nd convenience of drivers. Keywords: Technology cceptnce model, User experience, Autonomous vehicle, Advnced driver ssistnce system

2 88 Yujun Cho, et l. J Ergon Soc Kore 1. Introduction As digitliztion, multimedi, nd networking hve become core elements of utomotive technology, the development of intelligent high-technology vehicles is ccelerting (Yoon nd Kim, 2012). One of the core technologies for intelligent nd utonomous vehicles is the Advnced Driver Assistnce System (ADAS), which is driving-ssistnt device tht helps driver tke pproprite mesures or mintin sfe driving condition by controlling the vehicle utomticlly by mens of high-technology sensors nd intelligent imging units tht collect externl environment dt (Shin et l., 2016). The number of vehicles tht dopt the ADAS is incresing continully, nd the utomtion of specific driving elements such s speed control nd steering control re prticulrly dvncing (Rödel et l., 2014). The development of the ADAS, however, hs lso cused chnges in interctions between vehicle nd its driver (Eckoldt et l., 2012). While the decision-mking process ws crried out only by the driver in the pst, this system results in co-decision, tht is, coopertion between vehicle nd its driver (Rödel et l., 2014). For instnce, when it comes to decision to mintin certin level of speed, driver hs djusted the speed in recognition of the cr in front nd surrounding environments. Vehicles tht dopt the Advnced Cruise Control (ACC) function, s prt of the ADAS, djust the speed depending on the distnce from the cr in front utomticlly even if the driver does not step on the ccelertor. The driver my intervene only in certin sitution tht occurs. As such, interctions between cr nd its driver hve chnged. To gurntee pproprite interctions to drivers, therefore, the investigtion on the extent tht drivers re willing to give up prt of mnipultions in driving nd ccept utomtion of vehicles is of greter importnce thn before (Rödel et l., 2014). As for the reserch on drivers' cceptnce of the ADAS, the min focus hs been dditionl fctors to the existing Technology Acceptnce Model (TAM) model in order to explin the extent of cceptnce s in other similr res (Ghzizdeh et l., 2012). According to reserches on ADAS cceptnce, people, in generl, hve positive ttitudes towrd the ADAS nd utonomous vehicles, but Rödel et l. (2014) points out the limittion of existing studies in tht they neglect or fil to crefully tke into considertion the utomtion step of utonomous vehicles. Since this kind of utomtion my led of serious ccident chllenging to resolve when it involves n error, it is of significnt necessity to include the utomtion step s prt of reserches on the ADAS (Lee nd Kirlik, 2013). Prticulrly, driving conditions my involve mny thretening elements or irregulr situtions round the cr, nd thus ccepting the ADAS nd utonomous vehicles is quite sensitive mtter tht is directly connected with driver's life. Nonetheless, existing reserches re bsed merely on imginry scenrios of the ADAS s Rödel et l. (2014) pointed out. To collect ccurte dt regrding the extent of ADAS cceptnce, therefore, the necessity of cceptnce ssessment bsed on the utomtion step nd ctul experiences of users is more emphsized. Accordingly, this study includes cceptnce investigtion bsed on the utomtion step nd user experience, which hve been little considered in existing reserches on the ADAS, the core technology of utonomous vehicles. The objectives of the present study re s follows: First, it derives fctors tht ffect Behvbiorl Intention (BI) bsed on user experience ssessment by designing n cceptnce model of the ADAS, which is the core technology of utonomous driving nd by utilizing driving simultor; nd second, this study exmines chnges nd differences in ech fctor's User Experience nd Acceptnce (UX/UA) score depending on the utonomous vehicle utomtion step. 2. Relted Studies 2.1 Unified theory of cceptnce nd use of technology The TAM ws first suggested by Dvis (1986). The TAM is dvntgeous in tht it is simple nd esy to understnd, proving high level of explntory power mong vrious types of users nd systems (Im nd Oh, 2012). Since the limittion of the TAM ws Journl of the Ergonomics Society of Kore

3 30 Apr, 2017; 36(2): Technology Acceptnce Modeling bsed on User Experience for Autonomous Vehicles 89 therefter found in tht it filed supporting the vlidity of vrious extrneous vribles nd the reltion mong them, the developed Unified Theory of Acceptnce nd Use of Technology (UTAUT) integrted significnt 32 fctors from reserches on the use nd intention to derive 4 core elements. This integrtive theory lso presented n nlysis model on use intentions, cceptnce, nd use behviors (Venktesh et l., 2003). Reserches on the TAM dd new fctors to n existing model to explin the level of cceptnce mong users nd produced model thereof (Brookhuis nd Brown, 1992; Bjj nd Nidumolu, 1998; Dishw nd Strong, 1999; Krhnn et l., 2006). It is known tht for vehicles, the very object of this study, rework is essentil for technology cceptnce model in considertion of the specific condition - driving (Osswld et l., 2012). Accordingly, existing reserches tht include rework of the TAM nd UTAUT were exmined. For instnce, Adell (2010) mesured the extent of Driver Support System cceptnce by mens of UTAUT model, proposing modified UTAUT model bsed on the findings of the study on the effect of Performnce Expectncy (PE), Effort Expectncy (EE), nd Socil Influence (SI) on Behviorl Intention (BI) with existing question items of the UTAUT model modified. Osswld et l. (2012) conducted study with the im to propose driver cceptnce model in reltion to utomtion of vehicles. In ddition, the Cr Technology Acceptnce Reserch Model (CTAM) ws proposed bsed on recognized sfety nd nxiety fctors with n expnded UTAUT model. When the CTAM ws proposed (Figure 1), independent vribles included the bsic fctors of the UTAUT such s Performnce Expectncy (PE), Effort Expectncy (EE), Socil Influence (SI), nd Fcilitting Conditions (FC). In ddition, Anxiety (AX), Self-Efficcy (SE), nd Perceived Sfety (PS), Attitude towrds Using Technology (ATT) were lso included. Dependent vribles included Behviorl Intention (BI) nd Use Behvior (UB). This study exmines the extent of driver cceptnce of the ADAS. To derive vribles tht would ffect user behviorl intention, dopted ws n ITAIT model, which is the most successful integrtive model tht fetures superior explntory power to tht of the existing TAM nd includes mny vribles. Additionlly, the study model ws designed in reference to the CTAM nd on the bsis of the UTAUT model.

4 90 Yujun Cho, et l. J Ergon Soc Kore 2.2 Acceptnce of ADAS in the cr User cceptnce is defined s "n intention to use willingly informtion technology tht is embodied to support users (Dillon, 2001)". In ddition, the study of Adell (2010) defines driver cceptnce s "the extent of driver's s intention to use the system while driving if it is vilble". These definitions clrify the correltion between the extent of cceptnce nd use. In other words, n individul's intention of us is defined s driver cceptnce. As for existing reserches on ADAS cceptnce, Brookhuis nd De Wrd (2006) conducted study on cceptnce of utomted public trnsporttion vehicles (Philes). According to the result, it turned out tht drivers preferred complete utomtion for bout 55% of the entire route. Schller et l. (2008) exmined the cceptnce of single/combined versions of the ADAS in sitution of trffic congestion by mens of driving simultion. 81% of the prticipnts preferred the ACC function combining lne keeping ssistnce functions to single ACC functions. In ddition, bout 80% of the prticipnts showed high level of cceptnce with hnds off the steering wheel while using the combined ACC functions. Becker (1994) nd Nilsson (1995) s well showed tht prticipnts found the ACC useful, comfortble, nd sfer. While these reserches point out positive ttitudes towrd nd high level of cceptnce of the ADAS, there re studies on negtive spects. Helldin et l. (2013) visulized uncertin spects during scenrio of utonomous vehicle driving to verify its effect on the driver's Trust (T). The visuliztion of uncertin spects decresed the generl T nd cceptnce during the utonomous driving. De Wrd et l. (1999) conducted study on the driver's cceptnce in vehicle pltooning, nd it turned out tht the cceptnce of vehicle pltooning decresed s high level of risk nd low level of sfety were recognized. As such, reserches on ADAS cceptnce produced different results depending on the scope of study nd the level of utomtion. Additionlly, some reserches investigted ADAS cceptnce bsed on the fctors of User Acceptnce (UA) nd User Experience (UX). Adell (2010) mesured the extent of Driver Support System cceptnce by mens of UTAUT model, proposing modified UTAUT model bsed on the findings of the study on the effect of Performnce Expectncy (PE), Effort Expectncy (EE), nd Socil Influence (SI) on Behviorl Intention (BI). Trösterer et l. (2014) conducted study on the extent of cceptnce nd user experience over time s the prking ssistnce function of the ADAS ws utilized. Bsed on the existing CTAM thesis, the following UA nd UX fctors were selected: Performnce Expectncy (PE), Effort Expectncy (EE), Attitude towrds Using Technology (ATT), Socil Influence (SI), Fcilitting Conditions (FC), Self-Efficcy (SE), Anxiety AX, Perceived Sfety (PS), nd Behviorl Intention (BI). Trust (T) ws dded s one of the mjor fctors relted to ADAS functions. Rödel et l. (2014) clssified ADAS functions depending on the utomtion levels of Ntionl Highwy Trffic Sfety Administrtion (NHTSA) nd conducted study on the extent of cceptnce nd user experience bsed on the scenrio survey. UX/UA fctors included Perceived Esiness of Use (PEU), Attitude (ATT), Perceived Behviorl Control (PBC), Behviorl Intention (BI), Trust (T), nd Fun (F). Chnges in ech fctor depending on the utomtion step re exmined to derive mjor fctors nd verify the intention of using the ADAS. Unlike the study of Rödel et l. (2014) which ws conducted bsed on scenrio survey, this study utilizes simultor for ssessment bsed on user experience. In ddition, fctors tht might ffect BI were lso nlyzed. For this study, fctors tht might ffect BI, which indictes the extent of driver cceptnce, re selected bsed on literry investigtion. These fctors include PE, EE, SI, SE, AX, nd PS. Bnuls esed et l. (1997) conducted study on the correltion between driver AX nd the risk of ccidents. As result, it turned out tht AX ws in correltion with the risk of ccidents. In study of Groeger (1997), it lso turned out tht driver's sentiment (nxiety) ffected the driving performnce. In this study, ccordingly, it ws ssumed tht AX would ffect BI significntly. Perceived Sfety (PS) is mentioned s n importnt fctor of driving in study of Osswld et l. (2012). According to Alm nd Lindberg (2000), the perceived risk of ccidents ffect the driver's feeling of sfety nd nxiety. In this study, therefore, it ws ssumed tht PS would ffect BI significntly. Trust (T) is of specil importnce since it is relted to risks in driving. One study (Prsurmn et l., 2008; Trösterer et l., 2014) sttes tht the extent of T in utomtion is n importnt fctor tht decides the user's cceptnce nd relince. Thus, it ws ssumed tht T would ffect Journl of the Ergonomics Society of Kore

5 30 Apr, 2017; 36(2): Technology Acceptnce Modeling bsed on User Experience for Autonomous Vehicles 91 BI significntly nd tht chnges would occur depending on the utomtion step. Finlly, this study dds the fctor of Affective Stisfction (AS) which the user feels when experiencing the ADAS personlly during the driving simultion. Adell (2010) sttes tht in considertion of the specific sitution of vehicle, fctors through which the driver's sentimentl nd emotionl rections need to be dded to the UTAUT model. In existing reserches on informtion systems, user stisfction ws used s substitution vrible to system success. It is known tht this cn be n excellent indictor of chievement when system is used voluntrily t the user's discretion rther thn compulsorily s result of externl influence (DeLone nd McLen, 1992; Cvye, 1995). Thus, this study ssumes tht AS is n pproprite scle becuse the system is used voluntrily by the driver. In contrst, it turned out tht Fcilitting Conditions (FC) ffect UB rther thn BI directly in the UTAUT nd CTAM. Fcilitting Conditioins (FC) were excluded in this study becuse the focus is on the 3rd nd 4th levels of the utonomous vehicle driving utomtion which will be relized in the future. The dependent vrible is not UB tht is determined bsed on the current level of use nd use experience in the pst but BI tht indictes n intention or pln to use it t present or in the future. It is pproprite, therefore, to exclude FC tht is ssumed to ffect UB directly. Some existing studies (Wng et l., 2009; Lee, 2012) s well excluded UB nd FC in the nlysis. The ttitude towrd technology use (ATT) is defined s positive or negtive feeling nd n individul hs bout using certin technology. In this study, ATT is replced with Affective Stisfction (AS) which indictes the extent of emotionl/sentimentl experience tht n indivisul perceives while using technology. 3. Method 3.1 Study model nd vrible definition In this study, the study model nd mesurement items re selected bsed on fctors (PE, EE, SI, PS, AX, T, SF, SE, BI) tht were determined through the literry investigtion (Figure 2).

6 92 Yujun Cho, et l. J Ergon Soc Kore To exmine fctors tht would ffect BI, 8 independent vribles (PE, EE, SI, PS, AX, T, SF, SE) nd one dependent vrible (BI) were selected. Fctor definitions nd mesurement items re presented in Tble 1. To exmine chnges in UX/UA scores of ech fctor, the bsic scores of UX/UA of ech fctor were set s dependent vribles with utonomous vehicle utomtion s n independent vrible. The utonomous vehicle utomtion ws set to 4 levels: Level 1, Level 2, Level 3, nd Level 4. Additionlly, differences in UX/UA scores of ech fctor depending on pst uses nd sex distinction were exmined. For difference in UX/UA scores depending on pst uses, first of ll, UX/UA scores were set s dependent vrible with pst uses s n independent vrible. As for pst uses, if there is n experience of us, it ws set to Yes, nd otherwise, No. For difference in UX/UA scores depending on sex distinction, the bsic scores of UX/UA of ech fctor were set s dependent vribles with sex distinction s n independent vrible. Sex distinction ws set to 2 levels: men nd women. Tble 1. Item used in estimting the reserch model Vrible Fctor Definition Item Sources Performnce Expectncy (PE) Stndrd of driver's belief using ADAS function to help driving This function will be useful while driving This function will be importnt spect in cr sing this function is convenient to drive (Osswld et l., 2012) (Mdign et l., 2016) Effort Expectncy (EE) Stndrd of convenience wreness when using ADAS function It is difficult to lern how to use this function It is esy to become skillful bout the method of using this function Using this function is esy nd simple to understnd (Osswld et l., 2012) (Adell, 2010) (Bder et l., 2011) Socil Influence (SI) Stndrd of the importnce of other's opinion when using ADAS function Showing this function to others will mke me proud of myself Other people wnt me to use this function Other people will encourge me when I use this function (Osswld et l., 2012) (Kim nd Yoon, 2011) Independent vrible In the cse of lrge mount of time, I will implement this function perfectly Self-Efficcy (SE) Stndrd of user's belief in the bility using ADAS function Even though there is no one to let me know how to use it, I will be ble to implement this function perfectly (Osswld et l., 2012) If I hve n instruction of using this function, I will be ble to implement it perfectly Perceived Sfety (PS) Stndrd of ADAS function's influence on driver's helth nd hppiness I believe using this function might be dngerous Using this function does not relly bother me to drive I believe using this function is sfe (Osswld et l., 2012) Anxiety (AX) Stndrd of individuls' bility of perception nd response when feeling unesy I m worried bout using this function I believe this function might cuse ccidents I m not frid of using this function (Osswld et l., 2012) Journl of the Ergonomics Society of Kore

7 30 Apr, 2017; 36(2): Technology Acceptnce Modeling bsed on User Experience for Autonomous Vehicles 93 Tble 1. Item used in estimting the reserch model (Continued) Vrible Fctor Definition Item Sources Independent vrible Trust (T) Affective Stisfction (AS) Stndrd of positive expecttion of welloperted when using ADAS function with involving risk system Stndrd of positive, negtive, nd emotionl expecttion of driver fter using ADAS function I believe this function is verified professionlly I believe this function is relible I believe this function will be working in regrd to my expecttion This function mkes me stisfied Overll, this function meets my expecttion I will recommend this function to others (Ryu nd Kwon, 2005) (Prk et l., 2014) (Prk nd Kim, 2014) Dependent vrible Behvior Intention (BI) Stndrd of driver s willingness to use ADAS function Henceforth, I m willing to use this function in the future Henceforth, I m willing to use this function frequently when I will be ble to If it is possible to remin opertion continuously, I m willing to use this function consistently (Osswld et l., 2012) (Joo, 2013) 3.2 Assessment scenrio/function derived To derive ssessment scenrios depending on the utonomous vehicle utomtion step, this study refers to the 'Explntion on the 5 levels of Autonomous Vehicle Automtion' proposed by the NHTSA. According to the NHTSA (2013), Level 0 is the step of No-Automtion where mnipultions for driving re crried out solely by the driver. Level 1 is the step of function-specific utomtion where one or more mnipultions for driving re crried out utomticlly by the vehicle. Level 2 is the step of combined function utomtion where t lest two or more mnipultions for driving re crried out utomticlly by the vehicle. Level 3 is the step of limited self-driving utomtion where ll mnipultions for driving re crried out utomticlly by the vehicle within limited circumstnces. Level 4 is the step where every mnipultion is crried out utomticlly by the vehicle once the destintion is entered. This study derives 4 scenrios in which ADAS functions re utilized depending on the utonomous vehicle utomtion step. The scenrios consist of sub-sections: typing into the GPS system, driving in the city, nd driving on the express wy. Additionlly, uncertin situtions were provided t rndom to exmine how driver copes with n uncertin sitution nd tkes the responsibility. It ws ssumed tht experiencing such n uncertin sitution would ffect the extent of utomtion cceptnce significntly. Specific scenrios nd mjor functions derived in the NHTSA utonomous vehicle utomtion levels re presented in Tble 2 below. 3.3 Subjects As this study focuses on the 3rd nd 4th steps of the utonomous vehicle utomtion to be relized in the future, the ssessment ws conducted mong those currently in their 20s to 30s who would be in their 30s to 50s by the yer of 2030 when the technology would be mture enought to utilize these steps (Dokic et l., 2015; MLIT, 2015; Moon, 2013; Shnker et l., 2013). In reference to the ssessment criteri for NHTSA driving experience, the ssessment included drivers who would drive vehicle t lest 11,300km yer. The sex rtio ws bout 1:1-32 men nd 36 women - mong 68 subjects in totl (NHTSA, 2010). The verge ge of the subjects ws nd the stndrd devition ws

8 94 Yujun Cho, et l. J Ergon Soc Kore Tble 2. Description of the NHTSA levels of utonomous vehicle nd representtive function nd corresponding driving scenrio description NHTSA'S five level of utonomy Level 1 Level 2 Level 3 Level 4 Representtive function Cruise control or Lne keeping system Smrt cruise control + lne keeping system (Highwy driving ssist) Highwy utonomous driving system Full utonomous driving system Scenrio of evlution A Driver strts implementing nvigtion system nd input Kore University'. Following nvigtion s guidnce, driver should enter the highwy. After entering into highwy, driver should keep the speed limit 80~100km nd operte cruise control or Lne keeping system when he/she wnts. At this time, driver's eyes should be t the front nd driver's hnds should be holding hndle. In the mentime of using the function, driver should step on the brke when fcing emergency stte. After tht, driver cn deprt towrds destintion gin. A driver strts implementing nvigtion system nd input 'Kore University. Following nvigtion's direction, driver should enter the highwy. After entering into highwy, driver cn operte highwy driving ssist. At this time, driver's eyes should be t the front nd driver's hnds cn be plced off from hndle for five seconds. In the mentime of using the function, driver should step on the brke when fcing emergency stte. After tht, driver cn deprt towrds destintion gin. A driver strts implementing nvigtion system nd input 'Kore University' by using voice recognition. Following nvigtion's direction, driver should enter the highwy. After entering into highwy, driver cn operte highwy utonomous driving system with the guidnce of "this is where you cn strt utonomous driving" when he/she enters fixed intervl. As the system opertes, driving control will devolve on cr with the order of pedl., hndle, nd eyes. At this point, driver does not hve to look t the front nd cn off hnds from the hndle. In the mentime of using the function, driver should step on the brke when fcing emergency stte. If the intervl of utonomous driving is finished with the guidnce of "this is where to stop for utonomous driving", driving control will devolve on driver with the order of eyes, hndle, nd pedl. After tht, driver cn deprt towrds destintion gin. A driver strts implementing nvigtion system nd input 'Kore University' by using voice recognition. A driver will her "do you wnt to operte utonomous vehicle system?" from nvigtion. A driver cn operte full Autonomous driving system. As system is operted, every driving control will devolve on cr. At this point, driver does not hve to look t the front nd cn off hnds from the hndle. Like this, driver cn drive with utonomous vehicle system through ll the wy of highwy nd locl re. When there is n emergency, cr will give the lrm nd operte brke system itself to respond. After tht, system will be terminted with the guidnce of "you rrived t destintion" t the destintion. 3.4 Experiment environment The equipment prepred for this experiment consisted of the driving simultor, sitution bord, nd multi-touch monitor s shown in Figure 3. For the driving simultor, 3 units of 27-inch LCD monitors were used, nd the screen size of ech monitor ws 1,680 1,050. Thus, the totl size ws 5,040 1,050. Logitech G27 controller ws used with the steering wheel, pedls, nd gers s components. Additionlly, PNS GTs Plus wheel hngers nd sets were used. To simulte the driving environment of pssenger cr, EURO TRUCK SIMULATOR 2 developed by SOS Softwre ws utilized. For interctions with the driver while the ADAS ws functioning, 20-inch multi-touch monitor ws used. As for functionl opertion, pushes which re commonly used in moderndy utomobiles nd induce behviors of interctions between humn nd mchine were utilized (Byun, 2008; Bn, 2014). Journl of the Ergonomics Society of Kore

9 30 Apr, 2017; 36(2): Technology Acceptnce Modeling bsed on User Experience for Autonomous Vehicles 95 Specificlly, AXURE RP PRO 6.5 Softwre ws used to embody these switches. In ddition, for the sitution bord to disply current conditions of functionl opertion nd emergency situtions, 14-inch LG lptop computer ws utilized. The experiment mnger hd corresponding imges displyed on the screen. As for driving sections, destintion ws entered to the GPS nvigtion system with the cr stopped, nd the vehicle pssed the city, express wy, nd then rrived t the destintion. In the utonomous vehicle utomtion step, the experiment mnger ssisted by mens of keybord for driving nd other functions. 3.5 Experimentl procedures Interviews were conducted regrding mesurement items tht were determined bsed on literry investigtion, nd items found inpproprite were modified. Bsed on the modified set of mesurement items, the experiment ws conducted in utiliztion of the driving simultor. This experiment proceeded in the order of preprtion, prctice, nd ssessment. In the preprtion step, the experimentl objective nd procedures were explined to subjects. Descriptions no the historicl bckground of ech scenrio nd the ADAS were lso given. In the prctice step, bout 15 minutes were given to subjects so tht they could prctice mnipulting the controllers for the driving simultion prior to the ssessment step. Finlly in the ssessment step, the order of 4 ssessment scenrios ws rrnged in wy of Ltin squre for counter blncing in order to control extrneous vribles (Willims, 1949). Above-defined mesurement items were ssessed by mens of the 5-point Likert scle, nd interviews were conducted fter ech scenrio to collect subjective views. The experiment took bout 1 hour nd 15 minutes, nd n ssessment ws conducted once for ech scenrio with no repetition. 3.6 Anlysis method This study dopted SmrtPLS 3.0 (SmrtPLS GmbH, Bönningstedt, Germny) to test the study model. Bsiclly, the PLS nlysis involves no restrictions to the smple size s well s regulr distribution of residuls nd vribles. It is possible to nlyze the result only with smll number of smples. This method hs been verified s test method for predictive models in mny studies (Fornell nd Bookstein, 1982). To verify the sttisticl significnce of fctors tht might ffect BI, the nlysis of Smrt PLS Bootstrpping ws conducted. bootstrpping smpling ws implemented 1,000 times to verify fctors tht would ffect BI with the relince level s high s 95% nd the significnce level 0.05 respectively.

10 96 Yujun Cho, et l. J Ergon Soc Kore In this study, the verge of ech fctor's mesurement items utilized by Rödel et l. (2014) ws defined in UX/UA scores. To exmine differences in ech fctor's UX/UA scores, SPSS Softwre (SPSS 23.0, IBMSPSS, USA) ws utilized s the repetition mesurement vrince nlysis ws conducted for one vrible in the utonomous vehicle utomtion step nd the independent smple T-test ws conducted for the 2 vribles - pst uses nd sex distinction. 4. Results 4.1 Results regrding fctors tht ffect BI Study model test: relibility nd vlidity test To interpret nlysis results ppropritely, the relibility nd vlidity of the model nlysis results need to be confirmed. In generl, when Cronbch's lph coefficient is 0.6 or higher, the relibility is viewed s high enough (Nunnlly et l., 1967). In this study, the lph coefficient ws ll over 0.6, which indictes tht there is no problem in terms of relibility (Tble 3). Vlidity is relted to whether the trget of mesurement is ctully mesured. To judge construct vlidity in this study, convergent vlidity ws utilized. When the vlue of Averge Vrince Extrcted (AVE) is over 0.5 nd the vlue of Composite Relibility (CR) is over 0.7, it convergent vlidity is viewed s high enough (Chin, 1998). In this study, the AVE vlue ws ll over 0.5 nd the CR vlue over 0.7 respectively, which indictes tht there is no problem in terms of vlidity (Tble 4). Tble 3. Anlysis of relibility bout items Vrible Fctor The number of item Alph Performnce Expectncy (PE) Effort Expectncy (EE) Socil Influence (SI) Self-Efficcy (SE) Independent vrible Perceived Sfety (PS) Anxiety (AX) Trust (T) Affective Stisfction (AS) Dependent vrible Behvior Intention (BI) Tble 4. Anlysis of vlidity bout items Vrible Fctor Composite relibility AVE Performnce Expectncy (PE) Effort Expectncy (EE) Independent vrible Socil Influence (SI) Self-Efficcy (SE) Perceived Sfety (PS) Journl of the Ergonomics Society of Kore

11 30 Apr, 2017; 36(2): Technology Acceptnce Modeling bsed on User Experience for Autonomous Vehicles 97 Tble 4. Anlysis of vlidity bout items (Continued) Vrible Fctor Composite relibility AVE Anxiety (AX) Independent vrible Trust (T) Dependent vrible Affective Stisfction (AS) Behvior Intention (BI) Fctors tht ffect BI Derived vribles tht ffect BI include PE, SI, PS, AX, T, nd AS. EE nd SE re excluded from the list of vribles tht ffect BI. It turned out tht the explntory power of BI (R Squre Adjusted) ws quite high up to 0.84, which indictes tht PE, SI, PS, AX, T, nd AS explins BI s much s 84%. Specific results of the bootstrpping re presented in Tble 5. Tble 5. Anlysis of the bootstrpping method bout the pth coefficient Independent vrible Dependent vrible Smple men (M) Stndrd devition (STDEV) T p Performnce Expectncy (PE) ** Effort Expectncy (EE) Socil Influence (SI) ** Self-Efficcy (SE) Behvior Intention (BI) Perceived Sfety (PS) ** Anxiety (AX) ** Trust (T) ** Affective Stisfction (AS) ** (**: p-vlue < 0.05, *: p-vlue < 0.1) 4.2 Ech fctor's UX/UA scores Ech fctor's UX/UA scores depending on the utonomous vehicle utomtion step This study exmines chnges in ech fctor's UX/UA scores depending on the utonomous vehicle utomtion step in wy of the repeted mesurement vrince nlysis (Tble 6). Except EE nd SE regrding efforts tht were found not to ffect BI in the first result stted bove, the other fctors showed sttisticlly significnt difference depending on the utonomous vehicle utomtion step. In ddition, fctors tht re sttisticlly significnt depending on the utonomous vehicle utomtion step nd those tht re not re clssified nd presented in Figures 4 nd 5. As for UX/UA scores of fctors tht showed sttisticl chnges, the scores of SI, PS, T, nd AS except for PE incresed up to the 2nd step of utomtion, decresed down to the minimum in the 3rd step of utomtion, nd then incresed bit in the 4th step ccording to the Student-Newmn-Keuls (SNK) post-hoc test result (Tble 7). However, the tendency of AX,

12 98 Yujun Cho, et l. J Ergon Soc Kore which ws in n inverse proportion to BI ws the opposite to tht of other fctors. Tble 6. Result of ANOVA with repeted mesure design Fctor Levels of utomtion Stndrd F p Avg. (Independent vrible) devition SNK Post-hoc nlysis Performnce Expectncy (PE) ** ± 0.75 ± 0.67 ± 0.92 ± 0.77 b b Effort Expectncy (EE) ± 0.53 ± 0.53 ± 0.61 ± 0.79 Socil Influence (SI) * ± 0.83 ± 0.76 ± 1.02 ± 1.17 b Self-Efficcy (SE) ± 0.73 ± 0.77 ± 0.83 ± 0.97 Perceived Sfety (PS) ** ± 0.94 ± 0.79 ± 1.04 ± 1.01 b b Anxiety (AX) ** ± 1.16 ± 1.09 ± 1.29 ± 1.34 b b Trust (T) ** ± 1.05 ± 0.98 ± 1.11 ± 1.20 b b Affective Stisfction (AS) ** ± 0.88 ± 0.78 ± 1.01 ± 1.06 b Behvior Intention (BI) Totl verge of utonomous 4 levels 3.64 (**: p-vlue < 0.05, *: p-vlue < 0.1) Journl of the Ergonomics Society of Kore

13 30 Apr, 2017; 36(2): Technology Acceptnce Modeling bsed on User Experience for Autonomous Vehicles 99 Tble 7. Result of ANOVA with repeted mesure design (SI, PS, T, AS fctor in ccordnce with utonomous level) Levels of utomtion Mens Stndrd devition F p SNK Post-hoc nlysis ± 0.93 b ± ** ± 1.05 c ± 1.12 b (**: p-vlue < 0.05, *: p-vlue < 0.1)

14 100 Yujun Cho, et l. J Ergon Soc Kore Ech fctor's UX/UA scores depending on pst uses This study exmines differences in ech fctor's UX/UA scores depending on pst uses in wy of independent smple T-test (Tble 8). Tble 8. Result of t-test in ccordnce with pre-experiences Fctor Pre-experience (Independent vrible) T p Avg. Stndrd devition Performnce Expectncy (PE) Effort Expectncy (EE) Socil Influence (SI) Self-Efficcy (SE) Perceived Sfety (PS) Anxiety (AX) Trust (T) Affective Stisfction (AS) Behvior Intention (BI) Yes No ** Yes No 4.38 Yes No ** Yes No 4.09 Yes No ** Yes No ** Yes No ** Yes No ** Yes No ** ± 0.65 ± 0.90 ± 0.70 ± 0.56 ± 0.77 ± 1.05 ± 0.72 ± 0.90 ± 0.79 ± 1.05 ± 1.19 ± 1.32 ± 1.01 ± 1.16 ± 0.78 ± 1.03 ± 0.89 ± 1.18 (**: p-vlue < 0.05, *: p-vlue < 0.1) Except EE nd SE regrding efforts tht were found not to ffect BI in the first result stted bove, the other fctors showed sttisticlly significnt difference depending on the pst uses. Fctors tht showed sttisticl difference - PE, SI, PS, AX, T, nd AS - showed tendency of high UX/UA scores of fctors in the group with pst uses compred to in the group with no pst use. However, the tendency of AX, which ws in n inverse proportion to BI ws the opposite to tht of other fctors Ech fctor's UX/UA scores depending on sex distinction This study exmines differences in ech fctor's UX/UA scores depending on sex distinction in wy of independent smple T- test (Tble 9). SE tht ws found not to ffect BI in the first result stted bove showed sttisticlly significnt difference depending on sex distinction. Among fctors tht ffect BI, PS nd T showed sttisticlly significnt difference, nd the UX/UA scores of fctors were higher mong men thn women. However, the tendency of AX, which ws in n inverse proportion to BI ws the opposite to tht of other fctors. The rest of the fctors showed no sttisticl difference. Journl of the Ergonomics Society of Kore

15 30 Apr, 2017; 36(2): Technology Acceptnce Modeling bsed on User Experience for Autonomous Vehicles 101 Tble 9. Result of t-test in ccordnce with gender Fctor Gender (Independent vrible) T p Avg. Stndrd devition Performnce Expectncy (PE) Effort Expectncy (EE) Socil Influence (SI) Self-Efficcy (SE) Perceived Sfety (PS) Anxiety (AX) Trust (T) Affective Stisfction (AS) Behvior Intention (BI) Men Womn 4.06 Men Womn 4.34 Men Womn 3.40 Men Womn ** Men Womn ** Men Womn * Men Womn ** Men Womn 3.67 Men Womn 3.66 ± 0.74 ± 0.89 ± 0.67 ± 0.92 ± 0.84 ± 1.06 ± 0.67 ± 0.92 ± 0.86 ± 1.03 ± 1.17 ± 1.39 ± 0.88 ± 1.26 ± 0.87 ± 1.01 ± 0.99 ± 1.15 (**: p-vlue < 0.05, *: p-vlue < 0.1) 5. Discussion 5.1 Discussion on fctors tht ffect BI In the study of Adell (2010) on the cceptnce of the Driver Support System by mens of UTAUT model, the explntory power of BI ws s low s 20%. In contrst, the explntory power of the study model in this study ws s high s 84%. Two mjor cuses of such difference re considering user experiences by mens of driving simultor nd dding significnt fctors to the study model. In ddition to existing fctors such s PE, EE, nd SI, this study considers AX, PS, T, nd AS. As in Trösterer et l. (2014), Ghzizdeh et l. (2012), nd Helldin et l. (2013) reserches, it turned out tht T cused significnt effect on BI. T is n importnt element since driving is directly connected with life. The result ws significnt since the extent of T in the context of utomtion is vitl fctor tht decides drivers' cceptnce nd relince on utomtion. Thus, further reserches on T need to be conducted to find out wys to improve drivers' T. In ddition, it turned out tht s in De Wrd et l. (1999), Joshi et l. (2009) reserches, PS nd AX cused sttisticlly significnt effect on BI. As concerns bout sfety of the ADAS nd utomtion functions re rised due to the recent Tesl ccident, fctors relted to driver's sfety such s perceived sfety nd nxiety s well s bove-mentioned trust re emphsized. It is expected tht the importnce of driver sfety will increse s the utonomous vehicle utomtion step dvnces. Accordingly, relted compnies need to conduct ssessments of fctors relted to driver sfety in dvnce to ctul ppliction of the ADAS. Prticulrly when questionnire is used insted of driving simultor, rther conservtive pproch to the fctors will be necessry. It turned out tht AS showed lrge mount of T sttistics, indicting its significnt effect on BI. This mens tht AS my be used s substitute vrible to the ctul success of system. In this experiment, the driver ws free to use or stop using functions during the driving simultion. As it ws voluntry use t the driver's discretion

16 102 Yujun Cho, et l. J Ergon Soc Kore rther thn compulsory use, the condition ffected BI significntly. Hence, AS seems to be n importnt fctor to be considered in future reserches on the ADAS nd utonomous vehicles. In line with Rödel et l. (2014), Ghzizdeh et l. (2012), nd Adell (2010) reserches, however, EE cused no sttisticlly significnt effect on BI lthough it ws in contrst with the finding of Trösterer et l. (2014). As for ADAS cceptnce, results regrding EE my be different depending on model configurtion (difference in prmeters), ADAS functions, nd drivers' use environments (Ghzizdeh et l., 2012). In generl, EE on the efforts put forth through UTAUT Model cused significnt effect on BI when the user ws required to use computer progrm (Adell, 2010). In the cse of vehicles, however, secondry tsk ffects the driving performnce, nd thus lower lod nd less complicted behviors re required thn in the bove-mentioned cse of using computer progrm (Osswld et l., 2012). For instnce, if complicted process is required to implement function, it will ffect the primry driving tsk negtively nd increse the probbility of ccident occurrence. For this reson, current utomotive controllers dopt push switches tht re intuitive nd esy to use in order to minimize lod to the driver. In this study s well, push switches were used for every function in the utonomous vehicle utomtion step, nd thus EE cused no effect on the extent of cceptnce. In other words, s functionl implementtion of the ADAS needs to be intuitive nd esy in the contest of utomotive design, it turned out tht EE hd no effect on BI in this study. It lso turned out tht just s in the cse of EE, SE cused no effect on BI. 5.2 Discussion on the UX/UA scores of ech fctor Discussion on the UX/UA scores of ech fctor depending on the utonomous vehicle utomtion step Rödel et l. (2014) showed tht the UX/UA scores of PE decresed s the utonomous vehicle utomtion step proceeded, but in this study, the highest score ws in the 4th step of utomtion. One mjor cuse of such difference is considering user experiences by mens of driving simultor. As drivers experienced in the driving simultion tht s the utomtion step proceeded, personl behviors in driving decresed, they mrked higher scores on the usefulness nd convenience. Just s in previous reserches, T in this study decresed grdully s the utomtion step proceeded nd then somewht incresed in the 4th step. Throughout the utomtion steps, however, T scores were lower thn the verge of ech utomtion step (Tble 6). In generl, T dvnces with pst experiences of interctions. T in the ADAS s well is known to be ffected by drivers' personl experience of use (Rvld nd Grönroos, 1996; Trösterer et l., 2014). As in existing reserches, UX/UA scores of T depending on pst uses in this study show tht in the group with pst use experience UX/UA scores were higher thn those in the other group. Thus, it is expected tht s the use of the ADAS becomes common in the future, the score of T will increse. While in the study of Rödel et l. (2014), scores of most fctors in this study decresed s the utonomous vehicle utomtion step proceeded, scores incresed up to the 2nd step of utomtion, decresed nd then mintined or somewht incresed gin in the 3rd step. The lowest score ws shown in the 3rd step of utomtion probbly becuse, s in the study of Helldin et l. (2013), the process of hnding over the mnipultion right (conversion from mnul opertion to utonomous opertion or conversion from utonomous opertion to mnul opertion) cused nxiety nd hindered coopertion between the driver nd vehicle. Another cuse is tht the driver recognized the sttus s n immture step until it finlly becme complete utonomous driving. Thus, the future study needs to mke the process of hnding over the mnipultion right smooth when the 3rd step of utonomous vehicle utomtion nd prtil utonomous vehicle function re to be embodied. In the study of Rödel et l. (2014), when BI scores in ech utomtion step exceeded the verge of 4 steps of utomtion, it ws interpreted tht the driver hd n intention of use. Results of this study indicte tht the driver showed n intention of use in the 1st (Avg. 3.83) nd 2nd (Avg. 4.18) steps of utomtion, but in the 3rd (Avg. 3.43) nd 4th (Avg. 3.56) step, no intention of use ws Journl of the Ergonomics Society of Kore

17 30 Apr, 2017; 36(2): Technology Acceptnce Modeling bsed on User Experience for Autonomous Vehicles 103 displyed. These results re in contrst with the existing findings tht the intention of use ws confirmed ll in steps 1, 2, 3, nd 4 of utomtion nd tht subjects showed positive ttitudes towrd higher utomtion steps. Probbly, such difference results from experiences given to users by mens of the driving simultor, which were not noticeble through the scenrio questionnire. Drivers felt nxious lot during the response to emergency situtions in the 3rd nd 4th steps of utomtion. It lso turned out tht they felt nxious even while driving in interctions with other vehicles. Besides, they were frequently surprised t minor externl shocks while pying little ttention to driving fter hnding over the uthority to drive. However, the wy to judge the intention of use bsed on the verge score of the 4 steps of utomtion is dichotomous pproch tht is solely on the bsis of the score of single criterion. The future study needs to pply more logicl criteri for judgment. Since drivers' personl pst experiences of using the ADAS hve significnt effect nd the extent of cceptnce continues to chnge in the dvncement of technology, the study on cceptnce lso needs to be conducted continully (Kim nd Mlhotr, 2005; Trösterer et l., 2014) Discussion on the UX/UA scores of ech fctor depending on pst uses nd sex distinction As in the reserches of Rödel et l. (2014) nd Höltl nd Trommer (2013), it turned out tht every fctor tht ws found to ffect BI sttisticlly showed higher UX/UA scores in the group with pst use experiences thn in the other group. Thus, it is expected tht UX/UA scores of ech fctor will increse s the use of the ADAS becomes common nd the number of users increses. Among fctors tht ffect BI, PS nd T showed sttisticlly significnt difference, nd the UX/UA scores of fctors were higher mong men thn women. As for sex distinction, however, there needs to be further reserch becuse the scope of subjects ws limited in terms of ge nd scle. 6. Conclusion This study includes scenrios for ech step of utonomous vehicle utomtion nd ssessment of cceptnce bsed on user experiences nd in utiliztion of driving simultor. As for the result regrding the first objective, fctors tht ffect BI re sfety nd nxiety directly linked with life, including PS, AX, nd T s well s PE, SI, nd AS. Regrding the second objective, which is relted to fctors nd chnges in the dvncement of the utonomous vehicle utomtion step, UX/UA scores decresed in 3rd nd 4th steps of utomtion. Prticulrly in the 3rd step of utomtion, the process of hnding over the uthority in the conversion from/to the utonomous driving section cused nxiety, nd the driver recognized the process before the complete utonomous driving s immture. Besides, nxiety in the process of responding to minor shocks nd emergency situtions resulted in score decrese. UX/UA scores tended to be higher in the group with pst use experience thn in the other group, nd sfety-relted UX/UA scores were higher in the group of men thn in the group of women. It is expected, therefore, tht s functions re complemented nd their use becomes common, the extent of cceptnce will be better thn the current result. This study is of significnce regrding the following spects: First, this study presents n cceptnce model of the ADAS, which is core technology of utonomous vehicle driving, nd it verifies the model by mens of driving simultor tht cretes n environment where users cn experience the ADAS. It is expected tht the findings of this study will be bsis for future reserches on ADAS cceptnce. Second, chnges in UX/UA scores of ech fctor in ech step of utonomous vehicle utomtion were exmined depending on pst uses nd sex distinction. This result will be of help in developing pproprite ADAS elements nd grsping problems tht my ffect cceptnce in dvnce. Findings stted bove re expected to be useful for the development of tools to verify vlidity of functions prior to compny's product relese. It is lso expected tht the findings of this study will contribute to preventing potentil problems of pplying utonomous vehicle technology, developing technology tht users cn esily ccept, nd thus improving driver sfety nd convenience.

18 104 Yujun Cho, et l. J Ergon Soc Kore References Adell, E., Acceptnce of driver support systems, In Proceedings of the Europen Conference on Humn Centered Design for Intelligent Trnsport Systems, Berlin, Germny, , Alm, C. nd Lindberg, E., Perceived Risk, Feelings of sfety nd worry ssocited with different trvel modes, Pilot Study, KFB Stockholm: Linköping University, Bder, R., Siegmund, O. nd Woerndl, W., A study on user cceptnce of proctive in-vehicle recommender systems, In Proceedings of the 3rd Interntionl Conference on Automotive User Interfces nd Interctive Vehiculr Applictions, 47-54, Bjj, A. nd Nidumolu, S.R., A feedbck model to understnd informtion system usge, Informtion nd Mngement, 33(4), , Bn, K.M., Structurl modeling of ffective qulity for the opertion of utomotive switch, Kore University Grdute School, Bnuls esed, R., Crbonell Vy, E., Csonoves, M. nd Chisvert, M., Different emotionl responses in novice nd professionl drivers, Trffic nd trnsport psychology, Theory nd Appliction, , Becker, S., Summry of experience with utonomous intelligent cruise control (AICC), PART 2, Results nd conclusion, In World Congress on Applictions of Trnsport Telemtics nd Intelligent Vehicle-Highwy Systems, Towrds n Intelligent Trnsport System, 4, , Brookhuis, K. nd Brown, I., Ergonomics nd rod sfety, Ergonomics, 165, 35-40, Brookhuis, K. nd De Wrd, D., The consequences of utomtion for driver behvior nd cceptnce, Proceedings of the Interntionl Ergonomics Assocition (IEA), 10-14, Byun, J.H., Developing model of design for product user interction - Focused on generting informtion words through user observtion nd understnding interction elements -, Koren Society of Design Science, 21(1), , Cvye, A.L., User prticiption in system development revisited, Informtion nd Mngement, Informtion & Mngement, 28(5), , Chin, W.W., Commentry: Issues nd opinion on structurl eqution modeling, Dvis, F.D., A technology cceptnce model for empiriclly testing new end-user informtion systems: Theory nd results, Doctorl disserttion, Msschusetts Institute of Technology, De Wrd, D., vn der Hulst, M., Hoedemeker, M. nd Brookhuis, K.A., Driver behvior in n emergency sitution in the Automted Highwy System, Trnsporttion Humn Fctors, 1(1), 67-82, DeLone, W.H. nd McLen, E.R., Informtion systems success: The quest for the dependent vrible, Informtion Systems Reserch, 3(1), 60-95, Journl of the Ergonomics Society of Kore