Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/364178850

Avatar personalisation vs. privacy in a virtual try-on app for apparel shopping

Article in International Journal of Fashion Design Technology and Education · October 2022
DOI: 10.1080/17543266.2022.2125589

CITATIONS READS

2 557

3 authors:

Alex Ivanov Yi Mou


The Hong Kong Polytechnic University Shanghai Jiao Tong University
39 PUBLICATIONS 1,833 CITATIONS 58 PUBLICATIONS 1,150 CITATIONS

SEE PROFILE SEE PROFILE

Letwin Tawira
Shanghai Jiao Tong University
2 PUBLICATIONS 26 CITATIONS

SEE PROFILE

All content following this page was uploaded by Alex Ivanov on 11 October 2022.

The user has requested enhancement of the downloaded file.


International Journal of Fashion Design, Technology and
Education

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/tfdt20

Avatar personalisation vs. privacy in a virtual try-


on app for apparel shopping

Alex Ivanov, Yi Mou & Letwin Tawira

To cite this article: Alex Ivanov, Yi Mou & Letwin Tawira (2022): Avatar personalisation vs. privacy
in a virtual try-on app for apparel shopping, International Journal of Fashion Design, Technology
and Education, DOI: 10.1080/17543266.2022.2125589

To link to this article: https://doi.org/10.1080/17543266.2022.2125589

Published online: 04 Oct 2022.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=tfdt20
INTERNATIONAL JOURNAL OF FASHION DESIGN, TECHNOLOGY AND EDUCATION
https://doi.org/10.1080/17543266.2022.2125589

Avatar personalisation vs. privacy in a virtual try-on app for apparel shopping
Alex Ivanov , Yi Mou and Letwin Tawira
School of Media and Communication, Shanghai Jiao Tong University, Shanghai, People’s Republic of China

ABSTRACT ARTICLE HISTORY


Virtual try-on (VTO) apps are now used by many fashion consumers, but VTOs for the apparel Received 22 February 2022
category have met with resistance. This study examines privacy concern, body image and social Accepted 6 September 2022
value as antecedents to adoption intention towards an apparel VTO with two types of
KEYWORDS
photorealistic avatars. Twenty users first tried out the app in lab sessions, then 301 completed Virtual try-on; apparel retail;
an online survey with a video of the VTO. A majority of participants were concerned about privacy concern; fashion
potential misuse of their uploaded picture and preferred to use a pre-loaded avatar of a model consumer experience; body
with a similar body. This option explains why privacy concern had a weak negative impact on image
adoption intention in our model, albeit at the expense of self-presentation benefits. The trait of
privacy disposition best predicted consumer responses overall, yet other motives were also
revealed. Discussed are the implications of this study’s results and limitations to privacy calculus
research.

1. Introduction
Technologies Inc., 2021; Zeekit, 2021). Users may then
Billions of selfies (or ‘snaps’) are taken every day using proceed to purchase, save outfits on a wish-list, or
augmented reality (AR) filters in the Snapchat multime- share their new ‘look’ on social media.
dia messaging app (Accenture, 2022; Pierce, 2022). The Another reason for the lukewarm responses towards
fashion industry has been particularly eager to adopt AR apparel VTOs is related to body image. Some shoppers
technologies, and three-dimensional and hyper-realistic prefer to see garments displayed on professional models
clothing can now be ‘worn’ by your digital twin (Butler, rather than on their own ‘unflattering’ selves (Yang &
2022; Valmeo-Ogilvie, 2022). At the same time, one Xiong, 2019), and many others are concerned about
more basic type of AR interface that has been researched the privacy risk inherent in using AR mobile apps (Har-
for 20 years is still struggling to get acceptance by borth & Pape, 2021). But the Zeekit VTO may have
fashion consumers: the virtual try-on (VTO) app for solved this problem, as well, as it offers two avatar
apparel shopping on mobile devices. options. In addition to the high-personalisation option
VTOs enable us to see how a fashion product would mentioned above, a semi-personalised yet fully anon-
look on our real selves without physically trying it on ymous option lets users choose from almost 200 pre-
(Kim & Forsythe, 2008). Many such apps have been loaded avatars of non-professional models with various
well adopted for virtually trying-out beauty items or body types and sizes.
for in-store apparel shopping. Apparel VTOs for mobile Avatar personalisation in online apparel shopping
devices, however, have experienced lower user accep- has already been researched with a theoretical perspec-
tance (Riar, Korbel, Xi, Zarnekow, & Hamari, 2021). tive (e.g. Suh, Kim, & Suh, 2011) and a technological one
One reason is technological, as early prototypes disap- (e.g. Liu et al., 2020). Yet very few scholars have exam-
pointed users with crude avatars (Li & Xu, 2020; Plot- ined photorealistic visualisation of the tried-on clothing
kina & Saurel, 2019). Yet, since the pandemic a new in terms of privacy concern. Those who did so have
generation of VTOs has been launched that offers a focused either on cosmetics or accessories VTOs, or
much more accurate fit visualisation (Bhattarai, 2020; on hedonic use of AR in social media (Cowan, Javornik,
Lee, Xu, & Porterfield, 2020). The Zeekit and Forma & Jiang, 2021; Smink, van Reijmersdal, van Noort, &
apps available on the iOS store, for example, allow Neijens, 2020). Several studies (Merle, Senecal, & St-
users to mix and match thousands of tops and bottoms Onge, 2012; Zhang, Wang, Cao, & Wang, 2019) tested
on their photorealistic avatar, created in minutes based a research model for online apparel retailing, but not
on a picture taken with their smartphone (Forma in mobile or highly personalised contexts. Ours is the

CONTACT Yi Mou yimou@sjtu.edu.cn


© The Textile Institute and Informa UK Ltd 2022
2 A. IVANOV ET AL.

first study to examine how privacy concern and body less bearing on the purpose of our investigation. A per-
image factors influence user acceptance of an interface sonalised still picture can be good enough to assess fit,
with two modes of personalisation and/or privacy. We and it provides all the identifiability associated with
also contribute to the field by adding a social value con- privacy concern.
struct that can be instantiated with these new apparel
VTOs.
2.2. Pilot study
The specific focus on apparel can be conceptually and
socially justified. Context-specific privacy studies are 20 female undergraduates from a major university in
scarce (Xu, Dinev, Smith, & Hart, 2011) yet necessary China were recruited for the pilot study, consisting of
for engineering domain-tailored solutions (Cheah, 45-minute individual lab sessions. Participants first
Lim, Ting, Liu, & Quach, 2020). Apparel is also the lar- changed into tight-fitting clothing and took their
gest retail category worldwide, and better adoption of photo using an iPad Mini. The researchers then
VTOs could curb the economically and environmen- uploaded the photos to the Zeekit VTO installed on
tally-wasteful process of returned clothing (Cullinane, the tablet for avatar processing, and instructed partici-
Browne, Karlsson, & Wang, 2019). Improved accep- pants how to use the app. Participants were given 15
tance of VTOs could also inspire less fashion conscious min to browse, virtually try-on and select an outfit for
consumers to experiment more with their appearance a hypothetical date. The VTOs option of using the
and thus enhance their self-esteem. pre-loaded avatar was demonstrated next. (From hereon
For this research we adopted a lens for investigating we shall refer to the two avatar options as SelfAvt and
partially known-phenomena (Stebbins, 2001), starting OtherAvt.) The final phase was a semi-structured inter-
with an exploratory phase to inform our hypotheses view lasting ten to 15 min. Based on a thematic analysis
development and research design. This pilot study is of the transcripts (Peterson, 2017), we reached the fol-
described in the next section, after a brief review of lowing conclusions about VTO use, with representative
VTO literature. We then present the theoretical frame- quotes:
work and hypotheses, followed by the methods used for
the online survey with a video stimulus. The section . Of the 20, 11 participants had no privacy issues and
after that presents the quantitative results from the would definitely use SelfAvt. ‘Of course I prefer my
structural equation model as well as descriptive statistics own photo. I am buying the clothes for me so I
and qualitative data. The paper ends with theoretical have to try them on my body’. ‘I like using my
and practical implications, limitations, and photo. It will be more real and accurate. If the
recommendations. photo is private, why not?’
. Of the 20, 5 participants did have privacy concerns
and would therefore use OtherAvt. ‘I would use my
2. Related work own photo because even though body measure-
ments may be the same, details like hair style and
2.1. Virtual try-on
skin colour may not be. But because of privacy, I
The VTO stream of research began more than twenty would use a model’. ‘I’m not worried about the
years ago; early prototypes were desktop interfaces body. I am worried about my face because it is
with computer-generated avatars offering limited perso- exposed’.
nalisation. In Merle et al. (2012), for example, users . Of the 20, 4 participants would also choose OtherAvt
could modify the avatar’s body shape and skin colour but for non-privacy reasons. ‘Looking at my photo on
and upload an image of their head. In other studies ava- the app I look fat. Maybe I will use a model’. ‘Her
tars could be also rotated 360° to display fit from differ- photo is better than mine so I will use the model
ent angles (Fiore & Jin, 2003), yet it is debatable whether for better output choices’. ‘It takes less time than to
the VTOs in that generation offered AR. AR is different take my picture. I can find a model with similar
from virtual reality (VR), of course, in that the environ- body size’.
ment in VR is computer-generated, whereas AR super-
imposes digital content onto consumers’ real-world All participants reported the app to be useful and
(Javornik, 2016). Researchers have leaned towards easy to use, but most expressed doubts of actually adopt-
studying apps that work like a virtual mirror, such as ing it. Some disliked the clothing Zeekit offered and
the Sephora VTO (Scholz & Duffy, 2018), as the ‘live others were already satisfied with Chinese mobile shop-
AR’ experience that these VTOs offer, is still unavailable ping apps, such as Taobao. Most participants expressed
in Zeekit and Forma. This is a serious limitation but has having enjoyed the VTO’s Zeekit’s mix & match feature.
INTERNATIONAL JOURNAL OF FASHION DESIGN, TECHNOLOGY AND EDUCATION 3

2.3. Conceptual background spontaneous exploration to see new products and inno-
vations (J. Kang & Park-Poaps, 2011). Apparel VTOs
The research model for the main study draws on the-
may not be in a position to approximate the shared
ories of technology acceptance (Venkatesh, Morris,
hedonism in offline retail, but they do offer social
Davis, & Davis, 2003) and behaviour calculus (Laufer
value (SV) defined by Sweeney and Soutar (2001,
& Wolfe, 1977). Both frameworks posit consumer
p. 211) as ‘the utility derived from the product’s ability
decision-making as a calculus, or a trade-off between
to enhance social self-concept’.
two higher-order constructs: Perceived Benefits and
Similar value from using in-store or desktop VTOs
Perceived Costs. Benefits and costs in using a VTO
has already been shown by J.-Y. M. Kang (2014) and
were recently examined by Lee and Xu (2022); ours
Zhang et al. (2019) to positively impact perceived use-
focuses on costs relating to privacy and benefits from
fulness. To operationalise SV for our mobile context,
perceived social value.
we turn to the self-presentation and impression man-
Note, as well, that behavioural intention in studies of
agement benefits derived by social media users, as
technology acceptance typically measures system adop-
they regulate their personal image in the eyes of others
tion or purchase intention, while in privacy calculus
via selfies, videos, and likes (Rosenberg & Egbert, 2011).
studies (Culnan & Armstrong, 1999) intention is the
We suggest that the ability to easily generate share-able
willingness to disclose personal information, also called
previews of oneself wearing trendy clothing will be per-
self-disclosure. Users weigh perceived benefits (e.g. cus-
ceived as useful by mobile shoppers of apparel, and thus
tomised products, discounts, etc.) from such self-dis-
hypothesise:
closure against perceived costs (e.g. spam email,
identify theft, etc.). If they arrive at an estimate that H2. Social value will positively influence perceived
costs outweigh the benefits, users are more likely to usefulness.
maintain the status quo rather than adopt an innovation
(Dinev & Hart, 2006).
Figure 1 presents the key relationships and constructs 3.2. Privacy concern
(with greyed-out ovals showing constructs in our model).
Perceived Costs come in several varieties, as shown in
Omitted are Perceived Ease of Use and Enjoyment, as
Figure 1. Primary in technology acceptance studies is
these constructs capture process perceptions and ought
the effort users expect to exert in using the IT artefact
to be measured from actual use of a technology.
(i.e. perceived ease of use), but e-commerce also intro-
duces the risk of making a wrong assessment of product
3. Hypotheses development fit. Our scope here is limited to privacy risk or the risk of
opportunistic behaviour of a third party with access to
3.1. Usefulness and social value the disclosed personal information (Xu et al., 2011). If
The functional value provided to consumers is the key negative consequences of privacy loss matter to users,
determinant of technology adoption in general, and perceived risk becomes privacy concern (PC), which is
VTOs in particular are no exception (Lee & Xu, 2022). defined as ‘the degree to which an Internet user is con-
VTOs allow consumers to virtually try out a large num- cerned about web site practices related to the collection
ber of fashion items in a short time and at any place for and use of his or her personal information’ (Hong &
the purpose of assessing apparel size and fit. Thong, 2013, p. 276).
Such utilitarian value is traditionally measured by Users are especially concerned when a technology
perceived usefulness (USE), or the degree to which a requires access to their highly identifiable personal
user believes a technology will help them with reaching information, which includes their own photo (Mar-
goals (Venkatesh et al., 2003). USE has already posi- kos, Labrecque, & Milne, 2018). For the most part,
tively impacted adoption intention of VTOs in several however, it is the context that determines PC. Partici-
studies (Huang & Qin, 2011; Lee et al., 2020); confi- pants in Feng and Xie (2019) were concerned about
rming this relationship will serve as a benchmark for usage of their photos, but offering them more control
the strength of our hypotheses to follow. Hence we over the app’s privacy settings lowered their perceived
hypothesise: intrusiveness. Also perceived as intrusive was the
requesting of camera access in Smink et al. (2020)
H1. Perceived usefulness will have a strong positive yet this had no negative effect on their willingness
influence on adoption intention.
to allow it.
Fashion consumer journeys are not only about fulfil- No study has been conducted on VTOs requesting
ment of pre-planned shopping goals; they include access to a camera for capturing a full-body image of
4 A. IVANOV ET AL.

Figure 1. Theoretical framework.

the user, much less one wearing tight-fitting clothing. (2016) found that general online privacy concerns
Yet for users that bypass such self-disclosure we (akin to DP) influenced the ad-hoc, context-specific
expect reports of privacy concern to have only a assessment of privacy risk from a mobile shopping
weak negative influence on adoption. Overall, we task (akin to PC). No such study has been conducted
hypothesise: on apparel VTOs, which leads to the following
hypothesis:
H3. Privacy concern will have a weak negative influence
on adoption intention. H5. Disposition to value privacy will positively influ-
ence privacy concern.
With respect to PC’s effect on SV, however, we expect
a stronger negative effect. Not only would more private Finally, we examine the effects of body esteem (BE)
users be less inclined to share their avatars, they would on two of the mediators to adoption. Fashion consu-
be more likely to opt for OtherAvt, which would neu- mers often assess their appearance relative to other
tralise the potential for such value. Hence the consumers, or to other models (Rosa, Garbarino, &
hypothesis: Malter, 2006), yet only one study examined BE in
apparel VTO use. This was Merle et al. (2012), who
H4. Privacy concern will negatively influence social value.
found BE positively influenced confidence with apparel
fit. We also refer to Fan, Jiang, Deng, Dong, and Lin
3.3. Disposition to privacy and body esteem (2021), who found higher BE users to be more comfor-
table with uploading their selfies to social media than
Several studies examined the effects of individual traits lower BE counterparts, and state our final two
on VTO use. For instance, according to (Li & Xu, hypotheses:
2020), adoption intention is facilitated by higher levels
of fashion leadership and hedonic orientation. Given H6. Body esteem will negatively influence privacy
the body-centric lens of our investigation, we focus concern.
on disposition to value privacy (DP) and body esteem
H7. Body esteem will positively influence social value.
(BE).
DP is defined as ‘the individual’s need to shelter one’s The seven hypotheses, summarised in Figure 2, were
personal space or prohibit the disclosure of one’s private tested with a structural equation model of responses
information to others (Xu et al., 2011). Consumers with from an online survey. The survey also yielded descrip-
higher DP believe companies collect too much of their tive statistics and qualitative data that allowed us to
personal data for unspecified purposes (Liu, Wang, & determine which constructs predicted preferences for
Liu, 2019). Keith, Babb, Furner, Abdullat, and Lowry avatar personalisation.
INTERNATIONAL JOURNAL OF FASHION DESIGN, TECHNOLOGY AND EDUCATION 5

video is described in Figure 3 and can also be viewed


at: https://youtu.be/MeXVQiBQ3xM.

4.3. Measures
As shown in Table 1, scales were adapted from prior lit-
erature for all constructs except privacy concern, which
was developed for this study. Based on several com-
ments from the pilot, we included an indicator referring
to privacy concern with respect to one’s face. All indi-
cators were measured on a seven-point Likert scale.
Participants first rated their body esteem and disposi-
tion to value privacy, then proceeded to the video stimu-
Figure 2. Proposed research model.
lus. Only after it played in its entirety did they continue
with the survey, rating adoption intention, perceived use-
4. Method fulness, privacy concern, and social value, before being
4.1. Participants required to state their preference for either the SelfAvt
or OtherAvt option. The final question was open-
Our population consisted of consumers from the Gen- ended, asking for the reasons behind their choice.
eration Z demographic cohort, the most active mobile
shoppers of apparel (Shahbandeh, 2021). Participants
were recruited via the WJX.cn platform in China and 5. Results
compensated the equivalent of six US dollars for com-
5.1. The measurement model
pleting our survey. They had to be between 18 and 25
years of age, not to have used a VTO before, and to Cronbach’s alphas (α) shown in Table 1 are all above
have bought clothes online in the past year. We received .75, confirming reliability standards (Nunnally, 1978).
301 valid responses for analysis. Table 2 shows that the average variance extracted
(AVE) of all constructs is above .50, confirming discri-
minant validity (Hair, Ringle, & Sarstedt, 2011). The
4.2. Stimulus fit indices of the measurement model are satisfactory,
The video stimulus embedded in the survey was based on Hu and Bentler (1999): χ2/df = 2.832, p
produced with content from the Zeekit website < .001; TLI = .920 (>.90); IFI = .936 (>.90); CFI = .936
(http://zeekit.me) and screenshots generated in our (>.90); and RMSEA = .061 (<.08).
pilot sessions (with permission from participants).
As shown in Figure 3, emphasised were the mix &
5.2. Structural model results
match feature, avatar-creation process, social value
based on posting outfits on social media, and the two The structural model’s fit was tested with AMOS and
avatar personalisation/privacy options. The 40-second is also satisfactory (Hu & Bentler, 1999): χ2/df = 1.43,

Figure 3. Summary of the video stimulus.


6 A. IVANOV ET AL.

Table 1. Items and measurement model assessment.


Construct Source Items Loading α CR AVE
Body esteem (BE) Heatherton and Polivy (1991); Merle et al. (2012) Right now I feel satisfied with my body size and .89 .86 .92 .78
shape.
I am pleased with my appearance right now. .84
I feel satisfied with the way my body looks right .93
now.
Disposition to value Xu et al. (2011) It is most important for me to keep my online .87 .71 .87 .77
privacy (DP) privacy.
I am very cautious about posting selfies to social .86
media.
Compared to others, I tend to be more concerned .88
about threats to my privacy.
Adoption intention Lee et al. (2020); Venkatesh et al. (2003) When it is available I will download this app. .85 .76 .86 .67
(INT)
I intend to try this app for my clothes shopping. .80
I would recommend such an app to my friends and .82
family.
Perceived usefulness Plotkina and Saurel (2019); Venkatesh et al. The app is useful for trying-out clothes online. .83 .73 .85 .65
(USE) (2003)
The app would help me decide what clothes to .77
buy.
It would make my shopping for clothes more .82
effective.
Privacy concern (PC) Modified from Hong and Thong (2013) for the I’d worry that my personal information provided to .85 .85 .85 .68
photorealistic VTO context this app might be misused.
I’d be concerned about how private my uploaded .81
picture would be.
I will not feel comfortable uploading a picture of .84
my body to such an app.
I will not feel comfortable uploading a picture of .81
my face to such an app.
Social value (SV) J.-Y. M. Kang (2014) and Sweeney and Soutar Using the app would make a good impression on .86 .86 .91 .78
(2001) other people.
Using the app would improve my personal .90
relationships.
Using such an app would bring me social approval. .88

p = .22; TLI = .988 (>.90), IFI = .997 (>.90), CFI = .997 concern (PC) (β = .36, p < .001). H6, however, was
(>.90); RMSEA = .029 (<.08). rejected, as the path from body esteem (BE) to PC was
As shown in Figure 4, all hypotheses except H6 were non-significant. BE did have a significant positive influ-
supported. Not surprisingly the strongest correlation ence on SV, confirming H7 (β = 18, p < .001).
was that of perceived usefulness (USE) to adoption
intention (INT) (β = .68, p < .001). Also confirmed was
the hypotheses that USE would be significantly
influenced by social value (SV) (β = .36, p < .001). As 5.3. Descriptive statistics
for the effects of privacy concern, these were both nega- Of all 301 respondents, 75 (25%) chose SelfAvt vs. 226
tive towards INT and SV as hypothesised (β = –.09, p (75%) choosing OtherAvt. This three-to-one split indi-
< .05 and β = –.18, p < .001, respectively). PC negatively cates that the majority of our population prefer
influences INT indirectly via social value and perceived OtherAvt.
usefulness. These results differ from Huang and Qin We also compared construct means between SelfAvt
(2011) where perceived risk from a VTO neutralised and OtherAvt groups, as shown in Table 3. Participants
perceived benefits. Also confirmed was H5: disposition who chose OtherAvt had significantly higher DP and PC
to value privacy (DP) significantly influenced privacy scores than the SelfAvt group; the inverse was the case
for USE, SV, and INT. The groups did not significantly
Table 2. AVE values and squared correlations. differ in terms of BE.
BE DP INT PC SV USE Figure 5 also shows the variance of these means for
BE 0.88 the whole sample. The lower PC relative to USE and
DP −0.01 0.88 INT, alongside PC’s small effect sizes in the structural
INT 0.06 0.17 0.82
PC −0.04 0.34 −0.11 0.82
model, suggests that the privacy calculus of participants
SV 0.18 0.12 0.32 −0.13 0.88 favours perceived benefits over perceived costs from
USE 0.05 0.25 0.70 −0.03 0.35 0.81 privacy risk.
INTERNATIONAL JOURNAL OF FASHION DESIGN, TECHNOLOGY AND EDUCATION 7

Figure 4. Structural model results (**p < .05, ***p < .001).

In the open-ended question, we managed to extract


meaning from answers of 192 respondents. About
10% of responses included more than one motive, in
which case we counted the first one expressed. Motives Figure 5. Boxplot of construct means (n = 301).
behind SelfAvt preferences naturally revolved around
the realism of using one’s own picture, in line with the . OtherAvt Convenience: Just under 10 per cent pre-
pilot study. OtherAvt motives were less homogenous, fer to use a pre-loaded avatar due to the hassle in
yet also corroborated the sentiment of pilot participants, taking their own picture, worries if they are doing
with privacy being the main reasons to choose a pre- it correctly, etc.
loaded avatar. Seventeen per cent of participants cited
non-privacy reasons: more convenience or better
6. Discussion
image quality with the pre-loaded avatar. Exact figures
are given in Table 4. 6.1. Privacy vs. personalisation in apparel VTO
In sum, the preference split as and the motives in the use
larger sample echo those in the pilot sample, and the
The current study’s primary objective was to examine
findings can be summarised as follows:pt
the effects of privacy concern among Chinese females
.
between the ages of 18 and 25 towards using an apparel
SelfAvt Realism: about one-quarter of consumers
VTO with photorealistic avatars. In-depth interviews
prefer to use their own image as avatar; they find
were first conducted based on actual usage of such a
this more realistic and have no privacy concern.
.
VTO (the Zeekit app on iOS) by 20 participants, who
OtherAvt Privacy: Just over half of consumers would
took their own picture wearing tight-fitting clothing in
ideally use their own avatar, but choose not to due to
our lab prior to the task. We then produced a short
privacy concern.
.
video describing the key functionality of the VTO and
OtherAvt Body image: Just under 10% prefer to use
showed it to 301 participants with a similar demo-
a pre-loaded avatar as they expect better visualisation
graphic in an online survey. A structural equation
from the studio photography images or because they
model was used to test several hypotheses, and we also
believe the models’ bodies look better than their own.
looked at the descriptive statistics and qualitative data.
Concluded was that a marked majority of our popu-
Table 3. Construct means based on avatar preference. lation is concerned about potential misuse of an
Means for groups uploaded full-body picture of themselves, to the point
SelfAvt OtherAvt t- p-
Constructs (n = 75) (n = 226) Value Value
Privacy concern (PC) 4.43 5.40 −6.53 <.001 Table 4. Results from the open-ended question.
Privacy disposition (DP) 5.40 5.90 −4.30 <.001 Avatar choice & motive # of 192 %
Adoption intention 6.08 5.73 3.27 .001
(INT) SelfAvt Realism 49 26
Usefulness (USE) 6.10 5.79 2.86 .004 OtherAvt Privacy 110 57
Social value (SV) 4.33 3.87 2.85 .005 OtherAvt Body image 15 8
Body esteem (BE) 4.64 4.31 2.00 .047 OtherAvt Convenience 18 9
8 A. IVANOV ET AL.

that they prefer to use one of the VTO’s pre-loaded ava- concern was mentioned by several participants, although
tars. A minority of consumers would make the same not reflected in any of the BE indicators in the model. We
choice for reasons other than privacy, discussed in the therefore recommend future research to test a formative
next section. version of BE including this dimension, which appears
We further conclude that privacy concern has a weak relevant for AR contexts.
direct negative impact on adoption intention towards a We also recommend VTO technology acceptance
VTO, as long as an anonymous option providing some scholars to include effort expectancy with respect to ava-
personalisation is available. In other words, Zeekit’s tar creation. As for AR developers, we urge them to pay
OtherAvt feature appears to have neutralised the con- more attention to the image-capture phase in apparel
cern of individuals particularly predisposed to value VTOs, and to make sure the selection of pre-loaded ava-
privacy. It is less clear, however, to what extent their tars represents most body types of target users.
choice to use a semi-personalised avatar diminishes Finally, as Li and Xu (2020) have pointed out, it may
the benefits they would derive from a highly personal- be a long-term journey for VTOs to be accepted as an
ised avatar, had privacy not been an issue. Social value alternative solution to the physical fitting room. But
would certainly be sacrificed, since self-presentation is even now apps like Zeekit could revolutionise the
obviously tied to one’s own picture. We found that for mobile shopping experience for socially-connected
a sizeable portion of consumers the VTOs basic useful- and avid apparel shoppers who are glued to their smart-
ness is not compromised by the OtherAvt feature, as phone. As for consumers with a high disposition to
long as they can use avatars of similar body type and value privacy, even using a VTO without an anonymous
size. It is up to a future study, however, to accurately avatar option (like the Forma VTO) is not impossible, if
determine adoption intention towards VTO with only the app was part of a retail or fashion brand that these
one level of personalisation. consumers already trust. As a matter of fact, while we
were conducting this research, the Zeekit company
was acquired by the retailing giant Walmart, and
6.2. Social value and body image in VTO
Forma sold its VTO technology to Snap, Inc.
acceptance
Our research model confirmed the strong influence of
6.3. Limitations and implications for privacy
perceived usefulness (USE) on adoption intention and
calculus research
a medium effect from social value (SV) to USE. We
had conceptualised SV as self-presentation benefits A number of limitations are inevitable in studies such as
that consumers derive from using an apparel VTO. this one. A noteworthy methodological limitation is our
We urge fashion scholars to include such a construct single-factor research design: one stimulus describing
in their models. Our SV findings also have practical two different scenarios of personalisation/privacy. Our
implications in that VTO adoption could be increased initial plan was for a between-subjects experiment
if consumers are made aware of how easy it is to impress with two conditions, but a pre-test of the video treat-
friends on social media with apps like Zeekit and Forma. ments failed to induce differences in the constructs.
Based on the hypotheses testing and descriptive stat- Hence our decision to test the model with one con-
istics we conclude that body esteem (BE) does not play a dition. We strongly recommend, however, that future
key role in determining privacy concern or avatar pre- researchers aim for maximum variation between treat-
ference in apparel VTO use. These findings are odds ments of video interventions.
with studies like Merle et al. (2012), but not all research Also of note are the challenges with ensuring internal
reported positive effects from BE. Yim and Park (2019), validity in privacy calculus studies. When contrary
for instance, found users with lower BE had stronger beliefs and many unknowns are involved people tend
preferences for an eyewear AR feature than higher BE to rely on decision-making heuristics which favour the
users. Perhaps lower body esteem allows consumers to strength of one belief and override the other (Culnan
be impressed with their ‘augmented new body’, unlike & Armstrong, 1999). A user might have stated that
users higher on BE that may be habituated to favourable due to privacy concerns he or she would never use an
social comparison. app but would actually do so when faced with compel-
Body esteem did have a significant positive influence ling benefits (Kokolakis, 2017). Privacy assurances may
on social value, however, and body image as a general also backfire, as discovered by Wang, Duong, and Chen
notion does play a role in avatar preference. Many people (2016) when higher levels of privacy control activated
confident with their bodies may at the same time be self- users’ awareness of the probability of intrusion. It is
conscious as to how they come out in pictures. Such therefore difficult to make conclusions with regards to
INTERNATIONAL JOURNAL OF FASHION DESIGN, TECHNOLOGY AND EDUCATION 9

what kind of privacy calculus VTO users undergo, but Wells (Ed.), Contemporary operations and logistics (pp.
from the results and discussion so far it is clear there 301–322). Cham: Palgrave Macmillan.
is great variety in this process. Culnan, M. J., & Armstrong, P. K. (1999). Information privacy
concerns, procedural fairness, and impersonal trust: An
Overall, the study appears to confirm early e-com- empirical investigation. Organization Science, 10(1), 104–
merce findings by Chellappa and Sin (2005) that quan- 115. doi:10.1287/orsc.10.1.104
tified value of personalised services outweighs the Dinev, T., & Hart, P. (2006). An extended privacy calculus
quantified loss of privacy, yet consider how different model for e-commerce transactions. Information Systems
mobile shopping contexts are to those twenty years Research, 17(1), 61–80. doi:10.1287/isre.1060.0080
Fan, X., Jiang, X., Deng, N., Dong, X., & Lin, Y. (2021). Does
ago. Depending on the consumer segment and culture,
role conflict influence discontinuous usage intentions?
it could be that concern about face-related identify theft Privacy concerns, social media fatigue and self-esteem.
alone would preclude any consideration of VTO use. Information Technology & People, 34(3), 1152–1174.
Several of our pilot users stated such concerns, although doi:10.1108/ITP-08-2019-0416
they were not corroborated by the online survey respon- Feng, Y., & Xie, Q. (2019). Privacy concerns, perceived intru-
dents, nor did mean scores of the privacy concern indi- siveness, and privacy controls: An analysis of virtual try-on
apps. Journal of Interactive Advertising, 19(1), 43–57.
cator referring to one’s face differ significantly from the doi:10.1080/15252019.2018.1521317
one referring to one’s body. We would urge VTO Fiore, A. M., & Jin, H. (2003). Influence of image interactivity
researchers to investigate this further. on approach responses towards an online retailer. Internet
Research, 13(1). doi:10.1108/10662240310458369
Forma Technologies Inc. (2021). Forma closet virtual try-on
Disclosure statement (1.1.98). Cupertino, CA: Apple App Store. https://apps.
apple.com/ca/app/forma-closet/id1354631684
No potential conflict of interest was reported by the author(s). Hair, J. F., Ringle, C. M., & Sarstedt, M. (2011). PLS-SEM:
Indeed a silver bullet. Journal of Marketing Theory and
Practice, 19(2), 139–152. doi:10.2753/MTP1069-6679190202
ORCID Harborth, D., & Pape, S. (2021). Investigating privacy con-
cerns related to mobile augmented reality Apps – A vign-
Alex Ivanov http://orcid.org/0000-0002-7119-4880 ette based online experiment. Computers in Human
Behavior, 122. doi:10.1016/j.chb.2021.106833
Heatherton, T., & Polivy, J. (1991). Development and vali-
dation of a scale for measuring state self- esteem. Journal
References
of Personality and Social Psychology, 60, 895–910.
Accenture. (2022). Why the Future of Shopping is Social. Apple Hong, W., & Thong, J. Y. L. (2013). Internet privacy concerns:
Podcast. https://podcasts.apple.com/bg/podcast/built-for- An integrated conceptualization and four empirical studies.
change/id1561378348?i=1000568087906 MIS Quarterly, 37(1), 275–298. doi:10.25300/MISQ/2013/
Bhattarai, A. (2020, July 9). Virtual try-ons are replacing 37.1.12
fitting rooms during the pandemic. The Washington Post. Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes
https://www.washingtonpost.com/business/2020/07/09/ in covariance structure analysis: Conventional criteria ver-
virtual-try-ons-are-replacing-fitting-rooms-during- sus new alternatives. Structural Equation Modeling: A
pandemic/ Multidisciplinary Journal, 6(1), 1–55. doi:10.1080/
Butler, S. (2022, July 16). Buy now, try later: online clothing 10705519909540118
sales boom raises qualms over returns. The Guardian. Huang, N., & Qin, G. (2011). A study of online virtual fitting
https://www.theguardian.com/money/2022/jul/16/buy- room adoption based on UTAUT. 2011 International
now-try-later-online-clothing-sales-boom-raises-qualms- Conference on E-Business and E-Government, ICEE2011 -
over-returns Proceedings (pp. 1002–1005). doi:10.1109/ICEBEG.2011.
Cheah, J.-H., Lim, X.-J., Ting, H., Liu, Y., & Quach, S. (2020). 5881681
Are privacy concerns still relevant? Revisiting consumer Javornik, A. (2016). ‘It’s an illusion, but it looks real!’ Consumer
behaviour in omnichannel retailing. Journal of Retailing affective, cognitive and behavioural responses to augmented
and Consumer Services. doi:10.1016/j.jretconser.2020.102242 reality applications. Journal of Marketing Management, 32
Chellappa, R. K., & Sin, R. G. (2005). Personalization versus (9–10), 987–1011. doi:10.1080/0267257X.2016.1174726
Privacy: An Empirical Examination of the Online Kang, J.-Y. M. (2014). Augmented reality and motion capture
Consumer’s Dilemma. Information Technology and apparel e-shopping values and usage intention.
Management, 6(2–3), 181–202. doi:10.1007/s10799-005- International Journal of Clothing Science and Technology,
5879-y 26(6). doi:10.1108/IJCST-05-2013-0055
Cowan, K., Javornik, A., & Jiang, P. (2021). Privacy concerns Kang, J., & Park-Poaps, H. (2011). Motivational antecedents
when using augmented reality face filters? Explaining why of social shopping for fashion and its contribution to shop-
and when use avoidance occurs. Psychology & Marketing, ping satisfaction. Clothing and Textiles Research Journal, 29
38(10). doi:10.1002/mar.21576 (4), 331–347. doi:10.1177/0887302X11422443
Cullinane, S., Browne, M., Karlsson, E., & Wang, Y. (2019). Keith, M., Babb, J., Furner, C., Abdullat, A., & Lowry, P.
Retail clothing returns: A review of key issues. In P. (2016). Limited information and quick decisions:
10 A. IVANOV ET AL.

Consumer privacy calculus for mobile applications. AIS retailing. Journal of Retailing and Consumer Services, 51,
Transactions on Human-Computer Interaction, 8(3), 88– 362–377. doi:10.1016/j.jretconser.2019.07.002
130. doi:10.17705/1thci.00081 Riar, M., Korbel, J. J., Xi, N., Zarnekow, R., & Hamari, J.
Kim, J., & Forsythe, S. (2008). Adoption of virtual try-on tech- (2021). The use of augmented reality in retail: A review of
nology for online apparel shopping. Journal of Interactive literature. doi:10.24251/HICSS.2021.078
Marketing, 22(2), 45–59. doi:10.1002/dir.20113 Rosa, J. A., Garbarino, E. C., & Malter, A. J. (2006). Keeping
Kokolakis, S. (2017). Privacy attitudes and privacy behaviour: the body in mind: The influence of body esteem and
A review of current research on the privacy paradox body boundary aberration on consumer beliefs and pur-
phenomenon. Computers & Security, 64. doi:10.1016/j. chase intentions. Journal of Consumer Psychology, 16(1).
cose.2015.07.002 doi:10.1207/s15327663jcp1601_10
Laufer, R. S., & Wolfe, M. (1977). Privacy as a concept and a Rosenberg, J., & Egbert, N. (2011). Online impression man-
social issue: A multidimensional developmental theory. agement: Personality traits and concerns for secondary
Journal of Social Issues, 33(3), 22–42. doi:10.1111/j.1540- goals as predictors of self-presentation tactics on
4560.1977.tb01880.x Facebook. Journal of Computer-Mediated Communication,
Lee, H., & Xu, Y. (2022). Influence of motivational orien- 17(1), 1–18. doi:10.1111/j.1083-6101.2011.01560.x
tations on consumers’ adoption of virtual fitting rooms Scholz, J., & Duffy, K. (2018). We are at home: How augmen-
(VFRs): Moderating effects of fashion leadership and tech- ted reality reshapes mobile marketing and consumer-brand
nology visibility. International Journal of Fashion Design, relationships. Journal of Retailing and Consumer Services,
Technology and Education, 1–11. doi:10.1080/17543266. 44(June), 11–23. doi:10.1016/j.jretconser.2018.05.004
2022.2048423 Shahbandeh, M. (2021). Global apparel market - statistics &
Lee, H., Xu, Y., & Porterfield, A. (2020). Consumers’ adoption facts. https://www.statista.com/topics/5091/apparel-market-
of AR-based virtual fitting rooms: From the perspective of worldwide/
theory of interactive media effects. Journal of Fashion Smink, A. R., van Reijmersdal, E. A., van Noort, G., & Neijens,
Marketing and Management. doi:10.1108/JFMM-05-2019- P. C. (2020). Shopping in augmented reality: The effects of
0092 spatial presence, personalization and intrusiveness on app
Li, A., & Xu, Y. (2020). A study of Chinese consumers’ and brand responses. Journal of Business Research, 118.
adoption behaviour toward virtual fitting rooms. doi:10.1016/j.jbusres.2020.07.018
International Journal of Fashion Design, Technology and Stebbins, R. A. (2001). Exploratory research in the social
Education, 13(2), 140–149. doi:10.1080/17543266.2020. sciences. Thousand Oaks, CA: Sage Publications.
1758798 Suh, K.-S., Kim, H., & Suh, E. K. (2011). What if your avatar
Liu, Y., Liu, Y., Xu, S., Cheng, K., Masuko, S., & Tanaka, J. looks like you? Dual-congruity perspectives for avatar use.
(2020). Comparing VR- and AR-based try-on systems MIS Quarterly, 35(3), 711–729.
using personalized avatars. Electronics, 9(11). doi:10.3390/ Sweeney, J. C., & Soutar, G. N. (2001). Consumer perceived value:
electronics9111814 The development of a multiple item scale. Journal of Retailing,
Liu, Z., Wang, X., & Liu, J. (2019). How digital natives make 77(2), 203–220. doi:10.1016/S0022-4359(01)00041-0
their self-disclosure decisions: a cross-cultural comparison. Valmeo-Ogilvie, N. (2022, July 9). Augmented reality:
Information Technology & People, 32(3), doi:10.1108/ITP- Fashion’s newest frontier. ENVI. https://www.envimedia.
10-2017-0339 co/augmented-reality-fashions-newest-frontier/
Markos, E., Labrecque, L. I., & Milne, G. R. (2018). A new Venkatesh, V., Morris, M. G., Davis, G. B., & Davis, F. D. (2003).
information lens: the self-concept and exchange context User acceptance of information technology: Toward a unified
as a means to understand information sensitivity of anon- view. MIS Quarterly, 27(3), 425. doi:10.2307/30036540
ymous and personal identifying information. Journal of Wang, T., Duong, T. D., & Chen, C. C. (2016). Intention to
Interactive Marketing, 42, 46–62. doi:10.1016/j.intmar. disclose personal information via mobile applications: A
2018.01.004 privacy calculus perspective. International Journal of
Merle, A., Senecal, S., & St-Onge, A. (2012). Whether and Information Management, 36(4), 531–542. doi:10.1016/j.
how virtual try-on influences consumer responses to an ijinfomgt.2016.03.003
apparel web site. International Journal of Electronic Xu, H., Dinev, T., Smith, J., & Hart, P. (2011). Information
Commerce, 16(3), 41–64. doi:10.2753/JEC1086- privacy concerns: Linking individual perceptions with
4415160302 institutional privacy assurances. Journal of the Association
Nunnally, J. C. (1978). Psychometric theory (2nd ed.). New for Information Systems, 12(12). doi:10.17705/1jais.00281
York: McGraw-Hill. Yang, S., & Xiong, G. (2019). Try it on! Contingency effects of
Peterson, B. L. (2017). Thematic analysis/interpretive the- virtual fitting rooms. Journal of Management Information
matic analysis. In J. Matthes, C. S. Davis, R. F. Potter Systems, 36(3), 789–822. doi:10.1080/07421222.2019.1628894
(Eds.), The international encyclopedia of communication Yim, M. Y.-C., & Park, S.-Y. (2019). I am not satisfied with my
research methods (pp. 1–9). Hoboken, NJ: Wiley. doi:10. body, so I like augmented reality (AR). Journal of Business
1002/9781118901731.iecrm0249 Research, 100, 581–589. doi:10.1016/j.jbusres.2018.10.041
Pierce, D. (2022, April 28). Snapchat’s dress up feature turns Zeekit. (2021). Zeekit Virtual try-on (No. 2). Cupertino, CA:
your phone into an AR shopping mall. The Verge. https:// Apple App Store. https://zeekit.me
www.theverge.com/2022/4/28/23046385/snap-dress-up-ar- Zhang, T., Wang, W. Y. C., Cao, L., & Wang, Y. (2019). The
shopping role of virtual try-on technology in online purchase
Plotkina, D., & Saurel, H. (2019). Me or just like me? The role decision from consumers’ aspect. Internet Research, 29(3).
of virtual try-on and physical appearance in apparel M- doi:10.1108/IntR-12-2017-0540

View publication stats

You might also like