Learning From the Past to Improve the Future

The COVID-19 pandemic has created a state of emergency in countries around the world. In the early phase of the pandemic, governments considered contact tracing apps as one of the most promising tools to fight the virus and prevent lockdown measures. Corona-Warn-App in Germany, SwissCovid in Switzerland, and TousAntiCovid in France are just a few of the national apps that were developed and launched in 2020. Despite the high expectations, the adoption rates in most countries remained far below the threshold of 60% which corresponds to the desired percentage of a country’s population using contact tracing apps for them to be effective (University of Oxford 2020). In Europe, France with 47% (Rodgers 2021), UK with 40% (NHS 2021), and Germany with 39% (RKI 2021) boast the “best in class” adoption rates in 2021, although they still fail to meet expectations in terms of critical mass.

In many countries in the West, the introduction of contact tracing apps has been accompanied by controversial debates about their privacy implications and the risk of surveillance and revealed ethical or moral dilemmas (Rowe 2020). Accordingly, earlier research on contact tracing apps has mostly focused on the technology design for privacy-preserving apps (Ahmed et al. 2020; Cho et al. 2020; Yasaka et al. 2020). Because of the slow adoption rates, von Wyl et al. (2020) called for more research on the acceptability of contact tracing apps to provide an understanding of the rationale behind their use. Some of the first empirical studies that responded to this call are Trang et al. (2020), who analyze the impact of various app specifications (i.e., benefit appeal, privacy design, and convenience design) on app acceptance, as well as Meier et al. (2021) and Welrave et al. (2020), who examine users’ intentions to use contact tracing apps in light of privacy concerns. Other studies have analyzed how more users can adopt contact tracing apps. For example, Buder et al. (2020) apply choice experiments to assess users’ preferences for additional benefits, such as priority testing and food delivery, for increased uptake. Similarly, Jonker et al. (2020) assess the role of financial incentives on users’ adoption of contact tracing apps. Although these studies highlight that offering extended services can lead to a larger user base, their core focus remains on privacy design, which is reflected in the current app designs worldwide. Increasing adoption of contact tracing apps has proved to be a challenge, and involving users in the discussion on app characteristics and aspects related to the data processing is critical to ensuring mass acceptance (Redmiles 2020). This was also highlighted by Gupta and De Gasperis (2020), who suggest participatory design with users to help ensure that contact tracing apps meet their needs and are usable. While Trang et al. (2020) suggest there should be a one-size-fits-all app, their results also show that users’ preferences are far from uniform. In IS research, the privacy calculus explains the intention to use as the result of privacy trade-offs between expected benefits and perceived privacy risks (Dinev and Hart 2006). Thus, we anticipate an opportunity to address the varying preferences of different segments of the population with more targeted features that would provide both public and individual benefits. Wortmann et al. (2019) suggest that goal-congruent features, which are additional features on top of the core system functionality, can result in higher system use, regardless of its core features. For contact tracing apps, this implies that more attention is needed to provide services that offer benefits to users and can result in a larger user base.

This motivates our research goal to explore the role of goal-congruent features in improving the design of contact tracing apps and as an adoption driver. More specifically, we study the following question:

What are users' preferences for contact tracing app features, and what is the impact of value-added services on users’ adoption of these apps?

Based on a conjoint analysis (CA) study in Germany, we provide empirical insights into users’ preferences for core and privacy-preserving features, as well as value-added services of contact tracing apps. As an established market research technique, CA is a “practical set of methods for predicting consumer preferences for multi-attribute options in a wide variety of product and service contexts” (Green and Srinivasan 1978, p. 103). Occasionally, it has been used to study privacy trade-offs in the design of personal ICTs (Mihale-Wilson et al. 2017; Naous and Legner 2019). Our results confirm the dominant privacy-preserving design of most national contact tracing apps in Europe but also contribute to a more nuanced understanding of individual and group preferences. Following market simulations, we find that goal-congruent features – specifically, value-added services with a clear benefit structure – play an important role in driving user adoption.

From our study, we gain insights for a prescriptive design that allow the formation of app features that fit users’ expectations, with implications for service providers to adjust their offerings to different user segments. Our contributions are two-fold: first, our findings emphasize the role of goal-congruent features in addressing heterogeneous user segments with different benefit-risk tradeoffs and thereby fostering mass adoption. Second, methodologically we demonstrate that conjoint analysis – specifically, market simulation techniques – allow us to explore user preferences as a complementary method for participatory app design. Thus, our study contributes to the design efforts of contact tracing apps in particular, but more generally inform the user-oriented design of apps for digital health and crisis management that rely on sharing sensitive information.

The rest of the paper is structured as follows: First, we provide background on contact tracing in the context of COVID-19. Then, we introduce the applied research methodology, followed by a detailed description of the design of the CA. Next, we present the empirical results. Finally, we discuss our findings and conclude with implications for research and practice.

Our study builds on the idea of goal-congruent feature additions as drivers for the adoption of contact tracing apps. It aims to understand individual and group user preferences for different app designs and analyze the impact of value-added services on the adoption of contact tracing apps to improve their design.

We employ CA, which seeks to provide evidence of the factors that most influence the consumer’s choice of a product. Applying the utility concept from economics, CA enables the estimation of a user preference structure based on his evaluation of different product attributes or features. For these reasons, CA is gaining popularity to study information privacy trade-offs in different types of services (Krasnova et al. 2009; Ho et al. 2010) and is a very suitable method to inform IS design through an empirical analysis of user preferences.

In applying CA, we follow the methodological guidelines for IS studies outlined by Naous and Legner (2017, 2021) and use ACBCA, which extends the traditional full-profile CA (Green and Srinivasan 1978). This CA variant combines the advantages of adaptive- and choice-based procedures (Johnson et al. 2003). It is based on a choice experiment where participants have to choose from among a set of profiles (corresponding to different product combinations) after they perform a self-explicated task to assess must-have and unacceptable attribute levels from the evaluation to reduce the choice burden. Based on users’ choices, part-worth utilities and relative importance measures are calculated using the hierarchical Bayes (HB) estimation (Howell 2009).

The results from our conjoint analysis provide a micro perspective (i.e., that of the user) of users’ preferences for contact tracing apps through an evaluation of feasible design options. In contrast to prior research, our study goes beyond privacy-preserving aspects and the predominant black-box view of contact tracing apps. It provides a system evaluation by using a comprehensive set of features that include core and value-added services, as well as platform characteristics and user control (privacy-preserving features). Our approach helps improve app design and complements existing studies focused on user perception (see Table 2), thanks to the fine-grained assessment that highlights which of the features are required or most valued by users. By delivering data-driven insights that may serve as input for the participatory design of contact tracing apps (Gupta and De Gasperis 2020), our findings contribute to Pillar III of Von Wyl et al. (2020) regarding digital tracing apps.

With regard to individual user preferences, we find that exposure logging and test results sharing are the most important features in contact tracing apps, while exposure notification as the third core service lags far behind. Our findings support the dominant privacy-preserving design of most European contact tracing apps and confirm user preferences for a decentralized approach and contact tracing through proximity rather than location-based tracking via GPS. Despite the ongoing debate about privacy concerns and contact tracing apps, the results show that not all privacy-preserving features are valued by users. Previous research in IS has emphasized the negative impact of users’ privacy concerns on system use (Xu et al. 2009; Krasnova et al. 2010). Acquisti and Grossklags (2004) explain that users’ attitudes can be contradictory and result in a privacy paradox phenomenon. In this phenomenon, according to Barth and de Jong (2017), users are willing to compromise their privacy based on their assessment of the cost–benefit trade-offs. Accordingly, our results provide empirical evidence of privacy trade-offs with regard to contact tracing apps, in which users focus on the benefits associated with the app during a critical worldwide pandemic rather than the privacy risks or costs they entail. An alternative interpretation is that users trust these apps because they implement privacy through design principles and are implemented by authorities who protect the privacy rights of citizens through the EU General Data Protection Regulation (GDPR) (Yang et al. 2020).

Our study also provides interesting insights into the behavior of heterogeneous respondent groups with diverging preferences, represented by the user segments identified. Unlike Trang et al. (2020), we realize that no one app fits all and that different specifications of tracing apps contribute to their mass acceptance. While we observe a trend in favor of privacy-preserving features and basic functionalities (in the first and second segments), the largest user segment deems value-added services more important than privacy-enhancing features. The market simulation and variation analysis illustrate that contact tracing apps achieve higher market shares if value-added services are added beyond the basic app for tracing encounters. This is in line with research by Wortmann et al. (2019), who show that adding goal-congruent features to a core system may result in a higher rate of adoption. For contact tracing apps, the goal-congruent design implies a paradigm shift from a strong focus on privacy-aware design to explore value-added services that complement the app (e.g., through diagnosis and contextual services). Our results are in line with empirical evidence and current developments of the national contact tracing apps (see Table 1), where those countries that have integrated value-added services within the apps show an increase in the adoption rates. For instance, the TousAntiCOVID and the Corona-Warn-App have received several updates over time that included goal-congruent features such as a check-in service and vaccination certificate, which led to a greater user base. Hence, value-added services could become a game-changer in the adoption challenge.

To take privacy preferences into account, a viable implementation option is to integrate auxiliary apps with the COVID-19 app if needed. This is what Singapore did by merging the national TraceTogether app with the SafeEntry app as part of its safety measures to. The move led to an increase in the adoption rate (Lee 2020). A platform design could help avoid that several applications functioning in silos emerge, with none of them achieving critical mass: For example, Switzerland was the first country to promote a decentralized approach that preserves user privacy, but its app design did not have goal-congruent features until 2021, when it added an event check-in option. Individual apps are widely used in Switzerland, such as SocialPass (similar to the Luca app in Germany) for check-ins at restaurants and shops and a COVID Certificate app for vaccination. Applying a strategy similar to the one in Germany by including all these different features into the SwissCOVID app as a single platform can help increase user adoption to ensure broader protection while satisfying heterogeneous user needs.

In conclusion, our study contributes to the design efforts of contact tracing apps in particular and to design research in general. First, we contribute theoretically through emphasizing the role of goal-congruent features in addressing heterogeneous user groups and thereby fostering adoption. Building on the work of Wortmann et al. (2019), we argue that contact tracing app system design can be improved through the integration of value-added services that benefit the self as well as the public. Our results provide an illustrative example of the impact of goal-congruent features on system use and acceptability and help explain the uptake of contact tracing apps in Germany, France and UK. Second, we contribute methodologically by demonstrating the use of CA in exploring preferences with large, heterogeneous user groups and providing insights into trade-offs between core and privacy-preserving features as well as value-added services. Thereby, CA supports the participatory design and can complement existing methods through its different techniques. Specifically, market simulations bring insights into the feature selection and design process of these apps.

From a practical perspective, our results are relevant to the application developers and service providers of contact tracing apps. The preference model resulting from the CA study provides concrete realization options of the contact tracing app to be taken into consideration to gain sufficient critical mass and acceptability among users. Our findings support the development strategies of contact tracing apps, which have extended their services in the meantime with additional benefits (e.g., the German Corona-Warn-App has an integrated check-in function). To support their participatory design (Gupta and De Gasperis 2020), we provide a data-driven approach that allows capturing user preferences and including different stakeholders in the discussion of the most convenient design options. This approach is relevant for contact tracing apps, but can also inform the future design of mobile apps for health and crisis management that rely on sharing sensitive information (Behne et al. 2021).

Contact tracing apps have a national scope and, thus, may be impacted by both national implementation and contextual factors. As a result, an important limitation of our study is its focus on a sample from Germany, which has specific cultural characteristics and a democratic system, as well as an a priori model of decentralized contact tracing. It would be interesting to look at comparative studies in other countries that have different government regulations and app providers or have introduced centralized proximity or location-based tracking apps to assess the different design options. A cross-country perspective can provide additional insights into user trade-offs, which are governed by contextual and situational circumstances in their country.

Since privacy is a dominant topic in contact tracing apps, we have to acknowledge that our study focused on understanding user preferences for app design, but we did not use a detailed questionnaire to assess our respondents’ privacy awareness or concerns. Nonetheless, the results of the ACBCA clearly showed the intention to use value-adding instead of privacy-preserving features, which reflects the users’ a priori preferences and experiences.

Finally, it is important to note that our study took place while the contact tracing apps were being launched, and our selection of attributes was based on available and suggested options in June 2020. This implies that our CA study did not cover features that were created especially for testing and vaccination. However, the suggested methodological approach could be used to study those additional goal-congruent features in order to further improve the design of contact tracing apps and expand the use of market simulations in application design.