Risk preferences and risk perception affect the acceptance of digital contact tracing

The study investigates why acceptance and use of digital contact-tracing applications (DCTAs) remain low despite their potential to mitigate pandemics like COVID-19. Effective DCTA performance requires high adoption, yet Western populations show skepticism, often due to privacy concerns. The research compares individual factors (risk perceptions and risk preferences across health, economic, and data-security domains) with societal factors (social preferences, prosociality, identification with communities, trust) in predicting acceptance (willingness to use/recommend) and compliance (following DCTA recommendations after an exposure alert). The preregistered hypotheses posited that higher health-risk perception and lower data-security-risk perception would increase acceptance/compliance; risk tolerance would have domain-specific effects; and prosocial traits (e.g., honesty-humility, social value orientation, identification with broader communities) would positively predict acceptance/compliance.

Prior work links risk mitigation behaviors to knowledge and perceived risks during pandemics, with individual risk preferences shaping behavior despite similar perceptions. DCTAs have sparked debate about data-security risks, prompting privacy-preserving designs (e.g., decentralized Bluetooth-based proximity codes stored locally). From a societal perspective, DCTA adoption can be framed as cooperation in managing a common-pool resource (health system capacity), where social preferences and group identification may influence cooperation. Research shows prosociality predicts preventive behaviors, while heterogeneity and low trust can undermine cooperation. Thus, both individual risk-related factors and social preferences may affect DCTA acceptance and compliance.

Preregistered design and analyses (https://osf.io/b3ud5; materials/data https://osf.io/u6ngf/). Objectives: identify psychological factors influencing (i) acceptance of DCTAs and (ii) compliance with DCTA recommendations (e.g., self-isolation) following an alert. Participants: N=757 from a nationally representative German-speaking Swiss adult panel (LINK Institute), surveyed online in June 2020, one week after SwissCovid launch (initial N=848; 91 excluded for preregistered data-quality reasons). Mean age 45 (SD=16; range 18–79); 51.3% men, 48.3% women; 65% with at least high school diploma. Incentives: CHF 3.00 completion, and 10% random bonus tied to social value task decision. Measures: Acceptance (four items: inclination to use/recommend, perceived effectiveness, perceived data security; 1–5 Likert, averaged). Compliance (four items: willingness to self-isolate, report positive test, call hotline, get tested upon alert; 1–5 Likert, averaged). A standard vignette explained DCTA functionality. Social preferences: Social Value Orientation (six mini dictator games; angle measure), Honesty-Humility (HEXACO brief subscale), Identification with All Humanity (IWAH; world vs local identification). Risk preferences: SOEP general and domain-specific risk-taking items (0–10) for health, economic, general; plus added data-security risk preference item; z-standardized per participant. Risk perceptions: numeric estimates for past-7-day incidence, next-7-day incidence per 100,000, and severe-course rate per 100 infected; z-standardized across participants. Covariates: demographics (gender, age, income, wealth), DCTA comprehension (four-item scale incl. anonymity/data storage), technology affinity (four-item), and policy support for COVID-19 measures (four-item); multi-item scales averaged and z-standardized. Missing income/wealth imputed by sample medians (not preregistered). Analytic strategy: Bayesian linear regressions predicting acceptance and compliance from risk perceptions, risk preferences, social preferences, and selected covariates. All variables z-standardized. Covariate selection via Bayesian projective predictive model selection (reference full model; L1-penalized projection) to balance sparsity and predictive accuracy; risk and social variables retained as long as possible. Moderator analyses (preregistered) tested perceived DCTA effectiveness and local threat as moderators via Bayes factor comparisons.

Descriptives: 92% viewed COVID-19 as a global problem; 42% as a local/personal problem (36% no local problem). 1.7% reported a positive COVID-19 diagnosis; 11% self or close contact positive. Among those without infection/contact, 26% reported ≥1 COVID-like symptom. 93% worked from home fully or part-time; 15% reported income reduction. Policy support (1–5) was higher among women (M=4.09, SD=0.84) than men (M=3.86, SD=1.00) and higher in older (69–79 years: M=4.30, SD=0.87) versus youngest group (≤28 years: M≈3.86, SD=0.90). Risk knowledge: ~49% correctly knew deaths (~1,750) and 38% infections (~32,500) at the time; 97% identified chronic respiratory disease as a risk factor; 68% identified cancer; kidney disease was most misidentified (29%). 31% correctly stated past-7-day incidence; 44% overestimated; 27% correctly predicted next-7-day incidence; 51% overestimated; 50% expected no change. Acceptance and compliance: Acceptance mean 3.75/5 (SD=1.10); compliance mean 4.34/5 (SD=0.82). Acceptance lowest for ages 18–28 (M=3.41, SD=1.05) and highest for ≥69 (M=3.83, SD=0.93). 58% agreed SwissCovid was technically well designed; 89% believed insufficient population uptake. Model comparisons: Excluding social preference variables improved acceptance model fit strongly (BF≈20,271,851 vs full model), and outperformed models excluding risk perceptions (BF≈2,236) or risk preferences (BF≈12). No evidence for moderation by perceived effectiveness (BF≈7.94) or local threat (BF≈4.67). Acceptance predictors (standardized β, 95% HDI, BF): Strong positive effects of DCTA comprehension (β≈0.52, 0.46–0.58, BF>100) and policy support (β≈0.20, 0.14–0.26, BF>100); technology affinity also positive (as shown in figure). Risk perceptions: health risk perception positive (β≈0.05–0.06, HDI ~0.00–0.10, BF≈50); data-security risk perception negative (β≈−0.05, −0.11–0.00, BF≈37); economic risk perception negative (β≈−0.08, −0.14–−0.04, BF<1/100 for positive effect). Risk preferences: greater health risk aversion associated with higher acceptance (reflected as β for health risk tolerance ≈−0.06, −0.12–−0.01, BF≈50); greater data-security risk tolerance positive (β≈0.10–0.11, 0.06–0.15, BF>100); economic and general risk preferences showed no robust associations (economic β≈0.05, −0.01–0.11; general β≈0.01, −0.06–0.07). Social preference variables had small/negligible effects on acceptance (e.g., Honesty-Humility β≈0.04, 0.00–0.08). Compliance predictors: Compliance intentions were high overall. Risk perceptions had near-zero effects (CIs spanning 0). Risk preferences showed that greater health risk aversion predicted higher compliance (β for health risk tolerance ≈−0.08, −0.14–−0.02), and higher economic risk tolerance predicted higher compliance (β≈0.06, 0.01–0.11). The discussion notes more honest individuals were more likely to comply. Exploratory findings: Policy support increased with lower mental health (β≈−0.16), lower perceptions of data-security, economic, and general risks (β≈−0.14, −0.11, −0.12), and higher DCTA comprehension (β≈0.38). DCTA comprehension increased with higher policy support (β≈0.40), higher data-security risk tolerance (β≈0.11), and technology affinity (β≈0.10), and decreased for those not working from home (β≈−0.18) and with higher perceived data-security risk (β≈−0.14).

The findings address the central question of whether individual risk-related factors or social preferences better explain DCTA acceptance and compliance. Acceptance was primarily driven by individual considerations—perceived health risks of COVID-19, perceived data-security risks of DCTAs, and domain-specific risk tolerances—rather than social preferences. This suggests individuals weigh a trade-off between health protection and data-security concerns when deciding on DCTA use. Contrary to preregistered expectations, perceiving COVID-19 as an economic risk was associated with lower acceptance, potentially reflecting concerns about self-isolation’s economic impacts or a polarized focus on health vs economy. Beyond psychological risk factors, DCTA comprehension, policy support, and technology affinity had especially strong positive associations with acceptance, underscoring the importance of understanding and trust in governance. Compliance intentions were high and more strongly linked to individual risk aversion (particularly in health) and honesty, indicating willingness to follow recommendations upon alert is less of a barrier than initial uptake. Overall, enhancing comprehension and addressing specific risk perceptions, especially around data security, appear crucial for improving acceptance and, in turn, effective population-level use.

This preregistered, representative Swiss survey shows that acceptance of DCTAs is more strongly associated with individual risk perceptions and preferences, together with DCTA comprehension, policy support, and technology affinity, than with social preferences. Compliance intentions are high, but acceptance remains comparatively lower, limiting real-world impact. Practical recommendations include emphasizing personal health risks (including long COVID risks), clarifying the relatively low data-security risks of privacy-preserving DCTAs, and explaining how DCTAs function in accessible terms, highlighting individual and close-others’ benefits. Future research should experimentally test interventions to increase DCTA uptake, such as nudges, improved graphical risk/uncertainty communication, social comparison messages, and contextualizing DCTA data practices relative to familiar apps to reduce data-security concerns.