Privacy paradox in 3D body scanning technology: the effect of 3D virtual try-on experience in the relationship between privacy concerns and mobile app adoption intention

Three-dimensional body scanning has become accessible to consumers through mobile devices and is increasingly integrated with avatar-based virtual try-on (VTO) services for apparel fit and personalization. Despite potential benefits (accurate fit, mass customization), consumers must share detailed body images and measurements, raising significant body information privacy concerns that can hinder adoption. This study addresses: (a) the negative effect of consumers’ body information privacy concerns on their future intention to adopt mobile 3D body scanning technology, and (b) whether and how actual interaction with avatar-based VTO—via perceived control, personalization, and responsiveness—enhances consumer experience and mitigates privacy concerns to increase adoption. Guided by privacy calculus theory and the stimulus–organism–response (S-O-R) model, the research posits that interactive VTO features (stimuli) enhance consumer experience (organism), which in turn drives adoption intentions (response), potentially offsetting privacy concerns.

- VTO service types: photo-based, avatar-based, and augmented reality (AR)-based. Photo- and AR-based services are widely used for accessories/beauty and promotion but have limitations for accurate apparel fit. Avatar-based VTO allows users to input body measurements to personalize avatars and is particularly useful for online apparel fit assessment. - 3D body scanning: captures extensive body measurements to produce accurate avatars, promising better fit and reduced returns. Mobile approaches (smartphone cameras, LiDAR) are emerging, enabling at-home scanning, though reliability versus stationary scanners varies. - Privacy calculus and privacy paradox: Adoption of invasive, data-intensive technologies involves a trade-off between perceived benefits and privacy risks. For 3D scanning-based VTO, extensive body data collection amplifies privacy concerns (security, misuse, sharing), potentially deterring adoption. - S-O-R framework: Perceived interactivity—control, personalization, responsiveness—serves as stimuli that shape consumer experience (sensory, affective, intellectual), influencing behavioral intentions. Prior work shows interactivity enhances experience and intention, while privacy concerns increase perceived intrusiveness and reduce adoption. - Conceptual model and hypotheses: H1 posits privacy concerns negatively affect adoption intention. H2 posits perceived control, personalization, and responsiveness increase consumer experience. H3 posits consumer experience increases adoption intention. H4 posits consumer experience mediates effects of interactivity on adoption. H5 posits privacy concerns reduce perceived interactivity. H6 tests serial mediation whereby interactivity→experience mitigates the negative effect of privacy concerns on adoption.

Design: Online experiment providing actual interaction with an avatar-based VTO service (style.me demo). Participants created a personalized avatar using body information (height, weight, bra size, body shape), virtually tried on a gray jacket, adjusted size per recommendations, rotated the avatar, and submitted a screenshot to confirm engagement. Measures: After the VTO interaction, perceived interactivity—control (α=0.92), personalization (α=0.93), responsiveness (α=0.90)—from Wu and Wu (2006). Consumer experience measured as a second-order construct comprising sensory (α=0.92), affective (α=0.89), intellectual (α=0.95) experiences (Manthiou et al., 2016). Adoption intention for 3D body scanning technology (Kim & Forsythe, 2008). Body information privacy concerns (α=0.90; Roca et al., 2009). Five-point Likert scales used. Scenario for adoption intention: Participants were then presented with information about a retailer introducing a mobile 3D body scanning–based VTO to create an avatar from accurate scans (front/side images via AI-based calculation) and reported intentions to adopt. Sample and procedure: IRB-approved study. Pilot (n=32) on MTurk established instrument clarity and validity. Main sample collected on MTurk with high-quality worker criteria (>97% approval, >1000 tasks). Exclusions based on completion time (less than 10 minutes or over 1 hour). Final N=285 female participants; majority White, college-educated, aged 25–44, high incidence of online apparel purchasing; none had prior avatar-based VTO or mobile 3D scanning experience. Compensation: $1.50. Steps to reduce common method bias included Harman’s single-factor test and latent method factor controls; results indicated CMB not critical (variance ratio <50%). Analysis: Partial least squares structural equation modeling (PLS-SEM) with SmartPLS v3.3.3 tested three models: Model 1 (H1), Model 2 (H2–H4; S-O-R), and Model 3 (H1–H6; extended S-O-R with privacy calculus). Measurement model exhibited adequate reliability (loadings >0.60; CR>0.70), convergent validity (AVE>0.50), and discriminant validity (Fornell–Larcker, HTMT<0.90). Model fit acceptable (SRMR=0.07; NFI=0.90). R² values for endogenous constructs ranged approximately 0.26–0.62 across models. Bootstrapping used for significance testing.

- Model fit and explained variance: SRMR=0.07; NFI=0.90. R² for dependent variables across models ranged from 0.26 to 0.62. - H1 (Privacy concern → Adoption intention): Negative and significant. Model 1: β = −0.265, p<0.001. Model 3: β = −0.163, p<0.01. - H2 (Perceived interactivity → Consumer experience): All positive and significant. Control: β = 0.129, p<0.05; Personalization: β = 0.569, p<0.001; Responsiveness: β = 0.162, p<0.01. - H3 (Consumer experience → Adoption intention): Positive and significant. Model 2: β = 0.684, p<0.001; Model 3: β = 0.651, p<0.05. - H4 (Mediation of consumer experience between interactivity and adoption): Significant indirect effects. Control → Experience → Adoption: β = 0.088 (Model 2), β = 0.084 (Model 3), both p<0.05. Personalization → Experience → Adoption: β = 0.389 (Model 2), β = 0.370 (Model 3), both p<0.001. Responsiveness → Experience → Adoption: β = 0.111 (Model 2), β = 0.106 (Model 3), both p<0.01. - H5 (Privacy concerns → Perceived interactivity): Negative effects on Control (β = −0.219, p<0.001) and Responsiveness (β = −0.207, p<0.001); no significant effect on Personalization (β = −0.109, t=1.891). - H6 (Serial mediation: Privacy concern → Interactivity → Experience → Adoption): Partially supported. Significant only via Control: BPC → Control → Experience → Adoption: β = −0.018, p<0.05. Paths via Personalization (β = −0.004, t=1.729) and Responsiveness (β = −0.002, t=1.740) were not significant.

Findings confirm a privacy paradox in the 3D body scanning context: body information privacy concerns directly reduce intentions to adopt scanning-enabled VTO, yet interactive VTO features enhance consumer experience and, in turn, increase adoption intention. Among interactivity facets, personalization exerts the strongest positive effect on consumer experience, followed by responsiveness and control, underscoring the value of personally relevant, responsive interfaces in shaping sensory, affective, and intellectual engagement. Privacy concerns dampen perceptions of controllability and responsiveness, suggesting that heightened concern can impair users’ sense of system efficacy and timely feedback, which may further hinder adoption. Notably, privacy concerns did not diminish perceived personalization, highlighting personalization as a resilient design lever. The extended S-O-R model shows partial serial mediation: enhancing experience via control modestly reduces the negative impact of privacy concerns on adoption, whereas analogous serial effects via personalization and responsiveness were not significant statistically. Overall, the results indicate that well-designed interactive experiences can counterbalance, to a degree, privacy-related deterrents, thereby informing strategies to integrate 3D body scanning into VTO while addressing consumer concerns.

This study integrates privacy calculus with the S-O-R framework to explain adoption of 3D body scanning for avatar-based VTO. Empirical results show that body information privacy concerns hinder adoption intentions, while perceived interactivity—especially personalization—enhances consumer experience and increases adoption. Privacy concerns reduce perceptions of control and responsiveness but not personalization. Serial mediation is partially supported through the control → experience path, indicating interactive experiences can modestly mitigate privacy-related adoption barriers. Contributions include clarifying how distinct interactivity facets shape experience and adoption in privacy-sensitive VTO contexts and advancing understanding of experiential mechanisms that can offset perceived risks. Future research should examine additional experiential and risk-related factors (e.g., avatar appearance similarity, trust, transparency cues), user heterogeneity (age, culture, prior VTO use), and test design interventions (privacy assurances, data control features) in longitudinal or field settings, including diverse populations and real retail integrations.

- Self-report measures: Psychological experiences and intentions were captured via self-reports, which may introduce bias; common method variance was assessed and deemed acceptable, but experimental/behavioral measures would strengthen inference. - Model scope: A modest set of variables was examined; unmeasured factors (e.g., perceived avatar similarity to self, trust, transparency) may confound relationships. - Sample characteristics: MTurk sample of U.S. females may limit generalizability across genders, cultures, and markets; future studies should test demographic and cultural moderators. - Stimulus specificity: One VTO platform (style.me) and a single product category (a jacket) were used; results may vary across platforms, product types, and interface designs. - Cross-sectional design: Immediate post-exposure intentions were measured; longitudinal adoption and actual usage behaviors were not observed. - Technical/provider-side factors: The study did not model service-provider technical or psychological barriers that may influence integration and user perceptions.