INVESTIGATING INTERDEPENDENT PRIVACY ISSUES IN SOCIAL APP ADOPTION SCENARIOS: THEORETICAL RESULTS AND BEHAVIORAL EVIDENCE

Abstract

The popularity of third-party apps on social network sites and mobile networks increasingly highlights the problem of the interdependency of privacy. It is caused by users installing apps that often collect and potentially misuse the personal information of users’ friends who are typically not involved in the decision-making process.

We conduct two studies in the area of interdependent privacy to address the existing literature gap on this problem area and to work towards practical solution approaches. Motivated by the theory of other-regarding preferences, our research investigates to which degree users take their friends’ privacy into consideration when they make app adoption decisions.

In a first theoretical study, we provide an economic model and simulation results to investigate the adoption of social apps in a network where privacy consequences are interdependent. We present results from two simulations utilizing an underlying scale-free network topology to investigate users’ app adoption behaviors in both an early adoption phase and later adoption periods. The first simulation predictably shows that in the early adoption period, app adoption rates will increase when (1) the interdependent privacy harm caused by an app is lower, (2) installation cost decreases, or (3) network size increases. Surprisingly, we find from the second simulation that app rankings frequently will not accurately reflect the level of interdependent privacy harm when simultaneously considering the adoption results of multiple apps. Given that in the late adoption phase, users make their installation decisions mainly based on app rankings, the simulation results demonstrate that even rational actors who consider their peers’ well-being might adopt apps with significant interdependent privacy harms.

In the second study, we take an empirical approach to complement our theoretical work. Applying a conjoint study approach, we conduct the first study to quantify the monetary value which app users place on their friends’ privacy (i.e., value of interdependent privacy). Further, motivated by principles of contextual integrity, we examine the effect of data collection context on the valuation of interdependent privacy. We introduce two survey treatments: (T1) friends’ information is not relevant to app functionality, and (T2) friends’ information is relevant to app functionality. The results show that the monetary value (measured in US$) which individuals place on friends’ complete profile information is $1.56 in T1, and $0.98 in T2. In addition, we find individuals in T1 and T2 valuate their own complete profile information at $2.31 and $2.04, respectively. These valuations are significantly higher than the dollar values they place on friends’ complete profile information. We further measure the impact of comprehensiveness of data collection: an app may collect no information about users’ friends, basic information, or full profile information. Data collection context does not significantly affect how users value their friends’ basic information. However, regarding friends’ sensitive information (i.e., their complete profile), users in T1 valuate such information significantly higher than their counterparts in T2.

Both of these two studies contribute to the technology policy discussion on privacy in social apps by calling for meaningful market signals, e.g., designs that can better reflect the level of apps’ interdependent privacy harm, and mechanisms that inform users of apps’ data collection contexts. We believe such signals will help app users to make better informed decisions, and to more accurately address their own and their friends’ privacy preferences.

Chapter 1 |

Introduction

Over the last ten years, we have witnessed the rapidly increasing popularity of social network sites, with Facebook being the most successful entity. In order to expand its service and functionality, Facebook opened its platform to allow outside developers to interact with users through so-called third-party Facebook applications (or social apps). Similarly, the most important mobile platforms such as Android and iOS have enabled outside developers to create app content. These social apps gained worldwide popularity ever since their emergence. As of July 2013, Google Play, one of the largest mobile app platforms, facilitated 50 billion downloads of 1 million apps.^[1] Similarly, more than 75 billion times apps had been downloaded from another well-known mobile app platform, Apple Store, by June 2014.^[2] On Facebook, just to name one example, the Candy Crush Saga game app, is being used 10 million times daily.^[3]

Despite their high adoption rates, social apps pose growing privacy risks to users since they collect and potentially use users’ personal information. For example, since most users have little understanding of app permission management [2,3] , they tend to reveal more information to apps than they desire [2] . In addition, apps frequently collect more information than they need for their stated purposes, i.e., the apps are over-privileged [4,5] .

Perhaps even more troublesome, a newly addressed problem associated with app permissions is the interdependency of privacy, which refers to the phenomenon that in an interconnected setting, the privacy of individual users does not only depend on their own behaviors, but is also affected by the decisions of others [6] .^[4] The interdependent privacy issue is caused by users installing apps that often collect and potentially misuse the personal information of users’ friends, who typically have limited opportunities to grant explicit consent to or prevent these practices.

To date, only a limited number of research studies have started to address the problem of interdependent privacy. From an economic perspective, Biczók and Chia aim to define interdependent privacy and to provide initial evidence from the Facebook permission system for social apps. They further develop a game-theoretic model to analyze users’ app adoption decisions under the scenario of interdependent privacy. However, their study is limited to cases where two users are engaged in the decision-making over the adoption of one app, and therefore does not consider the complex dynamics of today’s app adoption behaviors.

Some other researchers investigate the interdependent privacy problem from non-economic perspectives. Biasiola presents a visual network analysis to discover the nature and scope of friends’ data leakage through users’ ties to an app [10] . In addition, through a case analysis of the Friendship Page on Facebook, Shi et al. discuss users’ concerns related to connectedness [11] . In a survey of Facebook users, Krasnova et al. [12] elicit levels of privacy concern regarding the release of 38 different information items including data about friends. None of these studies place individuals in a decision-making situation focusing on interdependent privacy as the central element of the investigation.

In order to address the literature gap mentioned above, we propose two studies in this thesis to investigate the problem space of interdependent privacy from both economic and empirical perspectives. Motivated by the theory of other-regarding preferences on how individuals care about their friends’ well-being [13] , our research investigates to which degree users take their friends’ privacy into consideration when they make app adoption decisions. In the first study, we follow an economic approach to study how large groups of users, who are connected in a complex social network, act in an interdependent privacy scenario. Mainly focusing on an empirical perspective, the second study applies a conjoint study approach to quantify the interdependent privacy concerns related to the adoption of a social app which may or may not collect information about users’ friends. In a nutshell, the first study presents theoretical results of social app users’ actions toward the interdependent privacy issue, while the results of the second study serve as behavioral evidence to support such theoretical discussion.

1.1 Overview of Two Studies

1.1.1      Study 1

In this study, we take a graph-theoretical approach and simulate app adoption decisions in scale-free networks to represent an approximate version of real social networks. More specifically, we conduct two simulations to investigate individuals’ app adoption behaviors in two phases. One phase is the start-up period of new apps, the other phase is the later app adoption stage. More precisely, the first simulation, which considers the iterative/sequential adoption process of social apps, is used to study users’ app adoption behaviors when an app is initially introduced. The second simulation, which is about comparing early adoption results of multiple apps, allows us to establish popularity rankings of the early adoption of those apps. We use those rankings to draw conclusions about the likely adoption processes of the considered apps in later adoption phases which are then heavily influenced by rankings [14] .

As expected, we find that in the initial adoption phase, app adoption rates will increase when (1) the interdependent privacy harm caused by an app is lower, (2) installation cost decreases, or (3) network size increases. In the second simulation, interestingly, we find that app rankings frequently will not accurately reflect the level of interdependent privacy harm when considering the adoption results of multiple apps. Our analysis implies that in the later adoption period, even rational actors who consider their peers’ well-being might adopt apps with invasive privacy practices. This helps us to explain why some apps that cause significant interdependent privacy issues are nevertheless highly popular on actual social network sites and mobile networks.

1.1.2      Study 2

Applying conjoint analysis, the second study allows us to estimate the monetary value of social app users’ interdependent privacy concerns. In addition, motivated by the principle of contextual integrity [15] , which suggests that privacy is contextdependent, we also aim to understand whether or not app data collection context influences how people care about others’ privacy. Specifically, we introduce two treatments (i.e., twoapp data collection contexts) into our survey design. In Treatment 1 (T1), friends’ information collected by the app does not improve its functionality; while in Treatment 2 (T2), the information the app collects about friends improves the functionality of that app.

In our study we distinguish between the collection of basic profile information and the collection of comprehensive profile information. Our results show that, on average, users in T1 and T2 would pay $1.56 and $0.98, respectively, in order to prevent an app from collecting comprehensive information profiles about their friends. Similarly, the dollar value an individual places on her own complete profile information when collected by a social app is $2.31 in T1, and $2.04 in T2. These results highlight that individuals take into consideration not only their own privacy, but also their friends’ privacy, therefore confirming the existence of other-regarding preferences in social app adoption scenarios. However, statistical analysis indicates that the dollar value participants place on their own privacy is significantly higher than what they place on friends’ privacy. These observations match the common-held belief that individuals care more about themselves than others.

When it comes to the app data collection context, on the one hand, we observe that it does not significantly affect how users valuate their friends’ basic profile information. On the other hand, users become more sensitive to app data collection context when an app tries to collect more valuable personal information about their friends, such as friends’ photos and locations (i.e., their complete profile). In other words, app users will increase their valuation of friends’ information when such information is not related to the functionality of an app and when it is detailed sensitive information.

1.2 Contributions

This thesis makes contributions to both the privacy literature and the technology policy discussion on privacy.

First, this thesis provides two studies to complement the still inadequately investigated research area of interdependent privacy. Specifically, taking an economic approach, the first study offers a theoretical discussion on social app users’ actions toward the interdependent privacy issue. The conjoint analysis conducted in our second study is the first attempt to quantify the value that social app users place on their friends’ information (i.e. value of interdependent privacy).

Second, the first study also complements the usable privacy and security studies which show that users install privacy-invasive apps because they are unable to identify and understand apps’ privacy consequences; however, we show that better informed and rational users will likely fall for privacy-invasive apps as well.

Third, our thesis contributes to the policy discussion on app privacy by calling for meaningful market signals. Our study results highlight the importance of introducing designs that can better reflect the level of apps’ interdependent privacy harm, and mechanisms that inform users of apps’ data collection contexts. We believe that such signals will help app users to make informed decisions, as well as to reveal their privacy preferences about both their own and their friends’ personal information.

1.3 Structure of Thesis

This thesis is structured as follows. In Chapter 2, we provide necessary background information about our study and discuss related work. Chapter 3 provides details about the app adoption model, simulation setup and simulation results in the first study. In Chapter 4, we provide details of the second study. Specifically, this chapter details our conjoint study design, experimental layouts, data analysis and empirical results. Finally, Chapter 5 concludes this thesis with a discussion about the implications of our study results and next steps for this strand of research.

[1] http://mashable.com/2013/07/24/google-play-1-million/

[2] http://www.statista.com/topics/1002/mobile-app-usage/

[3] http://www.socialbakers.com/facebook-applications/210831918949520-candy-crush-saga

[4] In the security context, several studies have considered the interdependency of decision-making, but those models are less applicable to the app adoption scenario [7,8] . For a survey of the results in the area of interdependent security see [9] .

INVESTIGATING INTERDEPENDENT PRIVACY ISSUES IN SOCIAL APP ADOPTION SCENARIOS: THEORETICAL RESULTS AND BEHAVIORAL EVIDENCE

Sharing is caring!