Default effects in app selection: German adolescents’ tendency to adhere to privacy or social relatedness features in smartphone apps

Abstract

Cognitive biases such as default effects impact on user preferences for a broad range of different choices. This paper investigates these default effects among adolescents configuring apps that either satisfy relatedness or enhance autonomy by protecting privacy. Relatedness and privacy are two innate needs that adolescents can satisfy with the use of smartphone apps. This study argues that adolescents’ choice of features supporting either privacy protection or social relatedness is a consequence of default effects, so that adolescents adhere to preselected defaults. We test this assumption in an experimental survey design including four app configuration tasks with N = 280 German adolescents aged 11 to 20 years. The study finds support for default effects for privacy as well as social relatedness. Effects for further variables, particularly age are discussed.

Keywords

adolescents apps default effects privacy relatedness

Smartphone ownership among adolescents has increased tremendously in recent years. For Germany, this study’s country of reference, it has increased among adolescents ages 12 to 19 from 47% in 2012 to 97% in 2017, and smartphones were the most prevalent way in which these adolescents accessed the Internet (Feierabend, Plankenhorn, & Rathgeb, 2017). At the same time, teenagers’ social lives often center on digital and mobile technology, mainly the smartphone and the use of its applications (apps) for a broad range of different functions.

Among these numerous functions, one in particular stands out. Adolescents use their smartphones not only to stay online but to (permanently) connect with their friends and peers (Vorderer, Krömer, & Schneider, 2016). To develop lasting social relationships with peers is one of the fundamental developmental tasks during adolescence (Peter & Valkenburg, 2011), and this task is more and more often being facilitated through the use of social media technology (boyd, 2014; Mascheroni & Ólafsson, 2014) on smartphones. Apps that facilitate social interactions between friends and peers on a local or global scale, such as Facebook, WhatsApp, and Snapchat, are among the most widely downloaded applications in the market (Feierabend et al., 2017; Jung, Kim, & Chan-Olmsted, 2014).

The need for social exchange and for the development of meaningful relationships with others is rooted in the basic motivational structures of humans as described in self-determination theory (Deci & Ryan, 2012). However, self-determination theory states that relatedness is not the only innate human need, but it also points to competence and autonomy as motivations for individuals. While smartphones may also be used to fulfill the need for competence through the availability of thousands of apps, autonomy becomes of particular relevance since its relationship with smartphone use can become a double-edged sword in terms of the satisfaction of this need. Smartphone apps not only facilitate social relatedness, but at the same time, they also challenge adolescents’ autonomy as they offer ample possibilities for infringements of adolescents’ privacy (Brown & Pecora, 2014; Madden, Lenhart, Cortesi, & Gasser, 2013). Apps give away personal data, and adolescents have to develop strategies to protect their privacy during mobile media use (Trepte et al., 2015).

In addition, research has demonstrated that people face challenges when protecting their online privacy (Acquisti, Brandimarte, & Loewenstein, 2015), particularly with respect to smartphones and their apps use (Egelman, Felt, & Wagner, 2013). For instance, social media users rarely set their privacy settings to align with their privacy preferences. Instead, they often accept the preset defaults (Stutzman, Gross, & Acquisti, 2013). Such behavior is rooted in what decision-making research describes as a status quo bias (Kahneman, Knetsch, & Thaler, 1991).

This study builds upon these cognitive biases for the valuation of certain app features to investigate how much default effects impact adolescents’ app preferences with respect to two central functions of apps: (a) enhancing their autonomy through privacy protection features and (b) developing social relatedness through social interaction and connection features. This line of questioning also leads to the issues of whether the default effects are stronger for autonomy and privacy or social relatedness, and which additional factors can explain adolescents’ preference for features supporting either one of these needs.

Empirically, this study relies on a research design that has already been applied with respect to default effects for privacy among adult smartphone users (Dogruel, Joeckel, & Vitak, 2017). However, this previous research lacked a comparison of autonomy needs with relatedness needs and did not address the specific importance that smartphones and apps can have in adolescents’ lives.

Theoretical basis: Self-determination theory and privacy

It is not surprising that smartphones and apps are attractive among adolescents as they fulfill crucial innate needs. According to self-determination theory, three driving forces of human behavior can be identified in relation to smartphone and app usage, which are the need for competence, the need for autonomy, and the need for relatedness (Deci & Ryan, 2000). People want to achieve mastery in what they do (competence), they want to have the feeling of being causal agents in their lives (autonomy), and they want to be socially connected with other human beings (relatedness). Digital media and smartphones in particular help satisfy these needs (Allen, Ryan, Gray, McInerney, & Waters, 2014). Certain app features specifically cater to the need for autonomy and relatedness. Autonomy is a central element for the conceptualization of privacy (Trepte & Reinecke, 2011), and relatedness can be achieved through the different functions that smartphone apps offer, which allow users to connect with friends, peers, or other larger communities.

Privacy can be considered a basic right for children and adolescents even when they are using media (Brown & Pecora, 2014). The concept itself relates to a broad range of disciplines from communication to sociology, psychology, and law (Margulis, 2011). From a communication scholar’s perspective, Westin (1967) provides a good starting point on which more recent approaches (Burgoon, 1982; Petronio, 2002) have built upon. He defines privacy as an individual’s control over what others know about him or her. For Westin, privacy has four functions: (a) personal autonomy, (b) emotional release, (c) self-evaluation, and (d) limited communication (Westin, 1967). Peter and Valkenburg (2011) demonstrated how these functions are closely linked to developmental tasks for adolescents. They stressed that having a sense of privacy allows an adolescent to develop into an autonomous individual, who has the competence to decide which information about himself/herself to share and which not to share. As such, successfully managing one’s privacy can be an important contribution to satisfying an adolescent’s innate need for autonomy (Deci & Ryan, 2000; Trepte & Reinecke, 2011).

Adolescents’ need for relatedness and how it can be met during their use of social media in general, and smartphone apps in particular, have been explored previously both theoretically (boyd, 2014) and empirically (Allen et al., 2014; Liu, Liu, & Wei, 2014). Adolescents use apps to connect with their friends and to develop meaningful social relationships. Self-disclosure, understood as the intentional revelation of information about one’s self to others, plays an integral role in the development and maintenance of personal relationships (Derlega, Metts, Petronio, & Margulis, 1993) and is closely linked to privacy. Privacy can be understood as managing what to self-disclose to others and what not to disclose, which is a central component of Petronio’s communication privacy management theory (Petronio, 2002) as well as the privacy calculus approach. Privacy or disclosure decisions are subject to a cost–benefit analysis such that individuals disclose personal information if they perceive that the overall benefits outbalance the assessed risks of disclosure (Laufer & Wolfe, 1977). While a great number of studies have addressed privacy behaviors based on the privacy calculus model (Dienlin & Metzger, 2016; Dinev & Hart, 2006), other approaches have departed from this rational approach by suggesting that people rarely engage in truly calculative decision-making but are prone to mental shortcuts or situational factors, such as affective states that impact the assessment of risks and benefits (Knijnenburg et al., 2017). This is a perspective this study applies as well.

A substantial amount of research has focused on either adolescents’ self-disclosure on social media, particularly Facebook (van Gool, van Ouytsel, Ponnet, & Walrave, 2015), or their privacy concerns (Feng & Xie, 2014; Youn, 2009) or on both aspects (Shin & Kang, 2016; Walrave, Vanwesenbeeck, & Heirman, 2012). However, most of this research has focused on self-disclosure and privacy in social media use in general, while not taking into account the specifics of mobile devices. Still, findings in this domain seem to be transferable to the context of self-disclosure and privacy on mobile devices as well. For instance, a general line of research argues that adolescents self-disclose more information online and are more hesitant to apply privacy protection strategies than adults (Taddicken, 2014; Walrave et al., 2012). This line of argumentation goes back to Barnes’s first description of the so-called privacy paradox (Barnes, 2006), typically described as an attitude–behavior gap between privacy attitudes and privacy behavior (Dienlin & Trepte, 2015). According to Barnes, this paradox is particularly accentuated among adolescent media users. At the same time, girls as compared to boys are more likely to be sensitive about potential privacy infringements (Madden et al., 2013; Youn, 2009).

Theoretical basis: App decision-making and cognitive biases

Research has started to focus on the role of media literacy as a method for fostering strategies to protect one’s privacy (Trepte et al., 2015). This stream of research argues that if individuals, and adolescents in particular, become more literate with respect to online privacy, they will be more likely to make decisions that protect their privacy. However, this perspective, at least implicitly, assumes that smartphone users would be not only sufficiently literate but also motivated to engage in extensive decision-making strategies, where cognitive defense mechanisms, such as media literacy (Rozendaal, Buijzen, & Valkenburg, 2009), come into play.

Yet, another line of research argues that decision-making for smartphone apps is often heuristically driven (Dogruel, Joeckel, & Bowman, 2015) and not so much a consequence of elaborated reasoning in which media literacy might guide users’ decision-making. As a consequence, app selection as well as privacy decisions related to smartphones and on social media are prone to cognitive biases during the decision-making process (Acquisti et al., 2015). For instance, Egelman et al. (2013) illustrated that a willingness to pay for privacy protection does not so much depend on the user’s privacy attitudes but on the way in which information about privacy is presented. With respect to privacy in general, Johnson, Bellman, and Lohse (2002) demonstrated that defaults (i.e., opt-out vs. opt-in) are successful in guiding users toward privacy-enhancement preferences on websites. Pointing in the same direction, Stutzman et al. (2013) determined that users do not make conscious decisions on what to disclose and what not to share on social media, but rather they tend to follow the preset defaults.

Such an adherence to default modes has been demonstrated as a stable cognitive bias in previous decision-making research and has been referred to as a status quo bias. Following this notion of a status quo bias (Kahneman et al., 1991), individuals use the status quo as a reference point and evaluate alternative options as advantageous or disadvantageous in their current situation. They are reluctant to switch from the alternative that they already use or own to another option (Bischoff, 2008). In economic terms, the willingness to accept a price for something like privacy is oftentimes higher than the individual’s willingness to pay for it. Dogruel et al. (2017) already demonstrated this with respect to default effects of privacy premium features for apps. In their study, defaults explained close to 9% of the preference for privacy premium features. Overall, defaults were able to increase the preference for privacy protection features by 40 to 50%.

Default effects, the reliance on the status quo, are rather stable and often assumed as medium-sized (Pichert & Katsikopoulos, 2008). They have been found to work in a broad range of areas (van Boven, Dunning, & Loewenstein, 2000). This study assumes that such default effects will have an impact not only with respect to privacy-related features but also with respect to social relatedness features. Default effects are independent of the object they address and should also be found among adolescents, regardless of their preference for privacy or self-disclosure.

Typically, most smartphone apps set social relatedness features as defaults, making it easy for users to connect with friends and peers. In contrast, privacy features that protect one’s autonomy are usually limited or only available for an additional cost. Therefore, it is interesting to examine how adolescents react to defaults when the situation is reversed.

Hypotheses

This study assumes that defaults are a potential way to increase individuals’ likelihood to consider privacy as a relevant feature in app selection (Dogruel et al., 2017). As such, this paper proposes:

H1: When adolescents are confronted with privacy protection defaults, they are more likely to keep such features even when they come at an additional monetary cost.

As already argued, default effects are also likely to occur with the other functions of smartphone apps that are central to adolescents’ life. As previously noted, social relatedness is an innate need for adults as well as adolescents, and a major driving force in self-disclosing information. This paper also proposes:

H2: When adolescents are confronted with a social relatedness default, they are more likely to keep such features even when they come at an additional monetary cost.

If the findings demonstrate default effects with respect to both privacy protection as well as social relatedness issues, the study will examine which of these effects are stronger. This paper hereby assumes that the default effects are stronger for the choices for which individuals have a preference. According to self-determination theory, however, both autonomy and social relatedness are considered crucial innate human needs and which of the two is more important to adolescents in the context of smartphone use remains unidentified. As a result, this paper proposes a research question comparing the strength of the default effects:

RQ1: For which domain—privacy or social relatedness—are these default effects stronger?

From a more explorative perspective, this paper investigates the extent to which other variables might influence the preference for privacy protection or self-relatedness features in smartphone apps. According to previous research related to privacy, sociodemographic as well as psychological factors are known to play a role in adolescents’ privacy attitudes and self-disclosing behavior. Typically, girls are more sensitive about privacy than boys, and privacy concerns are to a certain extent (Baruh, Secinti, & Cemalcilar, 2017; Dienlin & Trepte, 2015) a predictor of privacy protection. Age, as a developmental factor, and as a predictor and mediator of potential media effects (Valkenburg & Peter, 2013) may also impact the preference for both privacy protection as well as social relatedness features. The question arises about the extent to which privacy literacy might impact an individual’s preference for privacy protection and to what extent it may also mitigate cognitive biases, such as default effects. This study also examines the extent to which prior experience with apps plays an important role, which also taps into the domain of usage habit. Habitual media use, understood here as “a form of automaticity in media consumption” (LaRose, 2010, p. 194) without much cognitive control (LaRose & Eastin, 2004) may make users particularly prone to cognitive biases. Even though some theoretical assumptions on the potential effects of sociodemographic and psychological variables exist, these have not yet been tested with respect to cognitive biases, such as default effects. Therefore, this study opts for a more exploratory approach, carving out other potential relationships to be further investigated. This paper asks:

RQ2: To what extent is the selection for privacy protection features or social relatedness features a function of different sociodemographic factors as well as psychological variables?

Method

Participants

From February to March 2017 we recruited participants from a school in Germany (“Gymnasium”) as well as several youth clubs that collaborated with the researchers’ university. Students had to own and regularly use a smartphone to be eligible to take part in the study. We only recruited students attending secondary schools, resulting in an age range from 11 to 20 years. Participants did not receive compensation for their contributions, yet a nominal value (€0.50) was donated to the cooperating school’s support club for each student participating in the study. Two hundred and eighty students completed the questionnaire, which resulted in a statistical power of over 80% to identify medium-sized effects in an ANCOVA, as the study expected potential default effects based on the findings of Dogruel et al. (2017). The mean age of the participants was 14.49 years (SD = 1.76). The sample was biased towards female participants with 56% as compared to 33% being male participants; 11% of the participants did not identify as male or female or preferred not to answer the question. A majority of the participants (85%, n = 268) indicated their school type as a Gymnasium, the highest school form in Germany. Students had an average of 22.22 apps on their smartphones, but the variance was high (SD = 34.39, n = 274). The median number of apps was 14.

Procedure and materials

Adolescents were invited to take part in an online survey through their school or youth club. They were informed about the study 1 week prior, and their consent as well as their guardians’ informed consent, if deemed necessary by the school or participating youth club, were obtained on the questionnaire’s starting page. Within the online survey, the researchers conducted a one-factorial experimental design manipulating prerecommended app features, and replicated a research design previously employed by Dogruel et al. (2017). Following this strategy, the study employed an app configuration task for four app selection scenarios. Adolescents saw four apps in a row. Each app was presented with a generic description. The study employed newly generated app descriptions that were loosely based on existing apps and came from different domains that were considered attractive by adolescents in previous research (Feierabend et al., 2017). MeetMyDay was an organizer app that allowed for the creation and publication of events by inviting friends and coordinating leisure activities. FitNiceYo was a fitness app in which users could create their own workout schedule. WorldCraftGo was described as an augmented reality game where users could build virtual buildings in real-life environments. Music.me was a music app that scanned users’ music taste and recommended new music independent of platforms such as Spotify.

Again, replicating Dogruel et al.’s (2017) design, each app was complemented by four app features that could be selected or de-selected. These features were community support, function, personalization, and data use. For each of these four features, adolescents could choose between a basic and a premium mode. For example, in FitNiceYou, users could either train and track data for themselves (“community support basic”) or share and connect their workouts with friends (“community support premium”). In Music.Me, data were sent to a server unencrypted (“data use basic”) or encrypted (“data use premium”), as can be seen in Figure 1. For the manipulation of privacy defaults, the study relied on the data use feature. For the manipulation of social relatedness, the study relied on the community support feature. Each privacy premium feature highlighted either the ownership of one’s data or a secure way to share it. Each social relatedness premium feature highlighted the opportunities to connect with friends and/or a larger community.

Figure 1.

Example of stimulus MeetMyDay app (privacy premium default [Pdef] condition).

All apps were presented with a basic price of €1.00. Each premium feature selected added €0.50 to the app’s final price for a range of €1.00 to €3.00 per app. Adolescents first saw an app with preselected features according to the chosen research design and had to rate how much they would recommend this app to a friend. Then, they were given the opportunity to change the features according to their needs, and they had to make at least one change to the preselected version. The researchers randomly distributed the participants to one out of two conditions. In the privacy premium default (Pdef) condition, all four apps that the participants saw had the privacy premium feature selected and the social relatedness premium feature de-selected. In the social relatedness premium default (SRdef) condition, all four of the apps had the social relatedness premium feature selected and the privacy premium feature de-selected. Therefore, participants had either all social relatedness or privacy premium features as the four apps’ defaults. All other features—personalization and function—varied by app but remained consistent across the two experimental conditions. The researchers randomized the order of the apps, and participants had to configure and rate one app before they moved on to the next one (see Table 1). After configuring the app, they saw the reconfigured app again, including the new price, and they had to indicate how likely they would be to recommend the app again.

Table 1.

Stimuli presentation by experimental condition.

	MeetMyDay		FitNiceYo		WorldCraftGo		Music.Me
Pdef condition	Community	O	Community	O	Community	O	Community	O
	Function	*	Function	*	Function	O	Function	O
	Data use	*	Data use	*	Data use	*	Data use	*
	Personalization	*	Personalization	O	Personalization	*	Personalization	O
SRdef condition	Community	*	Community	*	Community	*	Community	*
	Function	*	Function	*	Function	O	Function	O
	Data use	O	Data use	O	Data use	O	Data use	O
	Personalization	*	Personalization	O	Personalization	*	Personalization	O

Note. Order of apps was randomized. Pdef = privacy premium default; SRdef = social relatedness premium default. *Premium feature set as default, O = basic feature set as default.

Measures

Privacy premium feature in configured app

For each app, this study measured if the final configured app had the privacy premium feature selected, either whether it was preselected (Pdef condition) or chosen (SRdef condition).

Social relatedness premium feature in configured app

For each app, the researchers measured if the final configured app had the social connectedness premium feature selected, either whether it was preset (SRdef condition) or chosen (Pdef condition).

Number of privacy premium features across tasks

As the selection or de-selection of a privacy premium feature could vary due to the preference for an individual app, the researchers averaged out the influence of individual apps by summing up the number of times privacy premium features were chosen for the configured app. This measure could range from 0 to 4.

Number of social relatedness premium features across tasks

Following the same logic, the researchers accounted for the number of times the configured app contained a premium social relatedness feature, again ranging from 0 to 4.

App evaluation

The study followed the design used by Dogruel et al. (2017) as it relied on a scale inspired by Reichheld’s (2003) research in regard to the net promoter score. Participants had to indicate how likely they would be to recommend the app to a friend who may need it, on a 7-point scale (1 = very unlikely, 7 = very likely). This was asked twice, one for the preconfigured app and one after it had been customized.

Privacy concerns

To account for privacy attitudes, the study relied on the Mobile Users’ Concerns for Information Privacy (MUIPC) Scale (Xu, Rosson, Gupta, & Carroll, 2012). Nine items measured perceived surveillance, perceived intrusion, and secondary use of personal information (M = 4.10, SD = 1.52) on a 5-point Likert-type scale (1 = fully disagree, 5 = fully agree). Internal consistency was high, with α = .902 (n = 280).

Habit strength of app download behavior

For measuring how common it was for the participants to download new apps, the study relied on an adjusted version of the Internet Habit Scale developed by LaRose and Eastin (2004; eight items; 1 = fully disagree, 6 = fully agree; M = 2.29, SD = 0.99) that had previously been employed for their research on adolescents’ smartphone use. Internal consistency was high, with α = .853 (n = 280).

Online privacy literacy (short version)

The Online Privacy Literacy Scale (OPLIS) is a tested measure to account for young people’s online literacy (Masur, Teutsch, & Trepte, 2017). The scale consists of four subdimensions: knowledge about institutional practices, knowledge about data protection laws, knowledge about technological aspects of data protection, and knowledge of data protection strategies. It employs a mix of true-or-false questions to statements about online privacy and multiple-choice questions (one true answer out of four). It consists of five items per subdimension for a total of 20 items. Since answering time is estimated to take around 15–20 minutes, the researchers had to shorten the scale. Therefore, the researchers selected two items for each of the four dimensions and included both true-or-false questions (five) and multiple-choice questions (three). Items were selected based on the item difficulty and to account for a broad range of privacy issues. As required, the number of correct items was summed up for an additive index (M = 2.88, SD = 1.83; n = 280).

Results

Controls and descriptive statistics

As the researchers had employed a randomization variable at the beginning of the questionnaire and not directly before the stimulus confrontation, dropouts and incomplete data led to a bias in terms of the participants being distributed to either two conditions. In the end, the study included 153 participants in the Pdef condition as compared to only 127 in the SRdef condition. However, conditions did not differ significantly with respect to age, t₂₆₇ = −1.21, p = .229; gender distribution, χ² (2, n = 250) = 0.583, p = .445; privacy concerns, t₂₇₈ = −0.45, p = .655; habit strength of app downloading behavior, t₂₇₈ = 0.52, p = .581; or online privacy literacy, t₂₇₈ = −1.25, p = .211.

For the first test, if the participants had successfully customized their apps according to their needs, the study tested if the preconfigured app offered was evaluated as worse than their own configuration. Therefore, the study employed paired t tests comparing the evaluation score for the preconfigured app and the later customized app. On average, participants evaluated their customized app significantly better (p < .01 all cases except Music.Me in the Pdef condition, here p = .024) than the preconfigured app. Consequently, the researchers considered the data suitable for testing the hypotheses.

The researchers relied on the descriptive data to examine how the participants had configured their apps. Table 2 gives an overview of how participants selected either a privacy premium feature or a social relatedness feature as a function of their experimental condition.

Table 2.

Frequencies for premium features selected by app.

	Privacy premium selected		Social connectedness premium selected
	Pdef condition(n = 153)	SRdef condition (n = 127)	Pdef condition(n = 153)	SRdef condition (n = 127)
MeetMyDay	65%	58%	36%	43%
FitNiceYo	63%	45%	26%	41%
WorldCraftGo	71%	52%	29%	32%
Music.me	65%	43%	21%	39%

Note. Pdef = privacy premium default; SRdef = social relatedness premium default.

On the descriptive level, two findings stand out. First, the default effect seems to have worked for both conditions, and it appears that it was stronger for privacy premium features than for social relatedness features. For all apps, more participants left the privacy premium feature active if it was preselected for them rather than selecting it actively when it was not presented as a default. The same was true for social relatedness features; however, here less than half of the participants kept the social relatedness premium feature if it was preselected for them. Without any further analysis, the descriptive findings seem to support H1 and seem to indicate a slightly weaker support for H2 as well. With respect to RQ1, the findings indicated a stronger effect with respect to privacy as compared to relatedness.

Hypothesis testing

To test hypotheses H1 and H2 and to answer the research questions, the researchers carried out two one-factorial ANCOVAs. For the first test, the number of privacy premium features across the four tasks acted as dependent variable (DV). For the second, the researchers relied on the number of social relatedness premium features across the four tasks. For the experimental manipulation, the specific default condition—Pdef or SRdef—was the main experimental factor. As covariates, the researchers accounted for age, gender (1 = male, 2 = female), privacy concerns, app downloading habits, and online privacy literacy (OPLIS).

For privacy premium defaults, the model used is significant and it explains a total of almost 11% variance. As expected in H1, experimental condition was a significant predictor. Those receiving the privacy premium default (M = 2.68; 95% CI [2.45, 2.91];¹ SD = 1.32; n = 139) chose significantly more privacy premium features than those in the SRdef condition (M = 1.97; 95% CI [1.68, 2.26]; SD = 1.49; n = 109; see Table 3).

Table 3.

ANCOVA statistics of number of privacy premium features selected.

	df	F (p)	Partial η²
Model	6	6.06 (p < .001)	.131
Intercept	1	0.10 (p = .749)	.000
Age	1	2.06 (p = .152)	.008
Gender	1	12.32 (p = .001)	.049
App habits	1	0.16 (p = .690)	.001
Privacy concerns	1	0.25 (p = .618)	.001
OPLIS	1	2.47 (p = .118)	.010
Experimental condition	1	13.587 (p < .001)	.053
Error	241

Note. OPLIS = Online Privacy Literacy Scale. Adjusted r² = .109.

With respect to social relatedness features, the findings indicated the same pattern as that for privacy premium features, yet to a weaker extent. As expected in H2, experimental condition was a significant predictor as well, so those presented with the social relatedness premium default (M = 1.57; 95% CI [1.29, 1.85]; SD = 1.44; n = 109) significantly chose more social relatedness premium features than those in the Pdef condition (M = 1.09; 95% CI [0.90, 1.27]; SD = 1.11; n = 139; see Table 4). Overall, the explained variance (adjusted r²) dropped to slightly under 4%.

Table 4.

ANCOVA statistics of number of social relatedness premium features selected.

	df	F (p)	Partial η²
Model	6	2.57 (p = .020)	.060
Intercept	1	2.62 (p = .107)	.011
Age	1	0.10 (p = .753)	.000
Gender	1	0.21 (p = .650)	.001
App habits	1	3.97 (p = .047)	.016
Privacy concerns	1	0.60 (p =.438)	.002
OPLIS	1	1.21 (p = .272)	.005
Experimental condition	1	8.25 (p = .004)	.033
Error	241

Note. OPLIS = Online Privacy Literacy Scale. Adjusted r² = .037.

This provides some evidence for answering RQ1. Adding to the descriptive analysis, the findings indicated a higher effect size (partial η²) for privacy premium defaults as compared to social relatedness defaults. However, a closer look at the data, particularly the descriptive data for the individual apps (see Table 2) as well as the combined level, shows that regardless of defaults, scores for social relatedness features were lower than those for privacy premium features. Regardless of the default condition, fewer participants chose social relatedness features as compared to privacy features. The question arises as to why this was the case. Since it could be a methodological artefact of the manipulation of the social relatedness feature, the researchers did not want to overestimate the absolute findings at this time. Instead, the researchers further examined the relative level. Privacy premiums as a default led to an average 36% increase in participants’ selection of privacy premium features. For social relatedness, having premium features as a default led to an average 44% increase. Thus, the default effect appears to be even stronger (on a relative basis) for social relatedness as compared to privacy premium features. What the study can conclude from this is that, considering all information, default effects appear to be similar for privacy and social relatedness, yet the overall baseline level was different, with the baseline for privacy premiums being higher than that for social relatedness, at least in this study.

Looking more into RQ2, the findings indicate a few significant covariates. The data show a significant effect for gender as it related to privacy premium features. This effect confirms the suggestion that, overall, girls (M = 2.61; 95% [2.40, 2.82]; SD = 1.40; n = 158) are more likely to choose privacy premium features than boys (M = 1.97; 95% CI [1.68, 2.28]; SD = 1.41; n = 92). This finding also aligns with previous findings that women and girls in general are more sensitive towards issues of privacy than boys (Madden et al., 2013; Youn, 2009).

For relatedness, this gender effect disappears. Here, a weak effect of app downloading habit strength is indicated. This effect runs in the direction that the more participants are used to download apps, the more likely they are to select social relatedness premium features. As relatedness is an inherent feature of many apps, this makes sense, which supports the assumption that for regular users of apps, relatedness becomes an even more important feature.

Exploratory data analysis for RQ2

For some further exploratory analysis, this study relied on a zero-order correlation matrix of all the variables employed in the study (see Table 5).

Table 5.

Zero-order correlation matrix.

	No. of social relatedness premiums selected (n = 280)	Age(n = 269)	Gender(n = 250)	Privacy concerns(n = 280)	App habits(n = 280)	OPLIS(n = 280)
No. of privacy premiums selected	r = .016[−0.113, 0.136]p = .784	r = .126[0.008, 0.237]p = .039	r = .218[0.092, 0.332]p = .001	r = .086 [−0.028, 0.203]p = .152	r = −.021[−0.144, 0.111]p = .725	r = .106[−0.003, 0.209]p = .078
No. of social rel. premiums selected	1	r = −.053 [−0.164, 0.054]p = .388	r = −.021 [−0.147, 0.102]p = .739	r = .013 [−0.117, 0.131]p = .833	r = .122[−0.004, 0.255]p = .041	r = −.077 [−0.184, 0.027]p = .199
Age		1	r = .010[−0.121, 0.129]p = .725	r = .099[−0.044, 0.240]p = .107	r = −.148 [−0.027, −0.258]p = .015	r = .309[0.196, 0.420]p < .001
Gender			1	r = .160[0.043, 0.27]p = .011	r = −.101[−0.222, 0.021]p = .110	r = −.173[−0.297, −0.038]p = .006
Privacy concerns				1	r = −.030[−0.166, 0.088]p = .612	r = .209[0.091, 0.319]p < .001
App habits					1	r = −.090[−0.205, 0.029]p = .134

Note. OPLIS = Online Privacy Literacy Scale. Results within square brackets are 95% confidence intervals (bootstrap with N = 1,000).

Based on this, the researchers noticed the observed effects of habit strength on the number of social relatedness features and that of gender on the number of privacy premium features that were selected. However, we also noticed that, not controlling for other variables, age and number of privacy premium features correlated positively with each other. The older the participants were, the more likely they were to choose more privacy premium features. The findings also showed two noteworthy and significant correlations with respect to OPLIS. First, a medium-sized correlation was found with respect to age, which can be interpreted as the older the participants were, the more they might have known about online privacy. Second, a weaker correlation was found in relation to privacy concerns, as the more privacy literate the users were, the more concerned they were about privacy (or vice versa). As a further step in the analysis, the researchers analyzed the extent to which age could moderate the effects of the experimental factors. We conducted a simple moderation analysis relying on the SPSS macro-PROCESS (Hayes, 2013) and chose gender, privacy concerns, app usage habits, and OPLIS as covariates. For the effect of the experimental condition on the number of privacy premiums, the findings did not show a significant moderation effect for age (t₂₆₉ = −1.85, p = .066). The conditional effect was significant for participants whose age was below 1 SD (p <.001) and for those within 1 SD from the mean (p < .001), but not for those 1 SD above the mean (p = .193). For the effect on the number of social relatedness premiums chosen, the moderation effect was also not significant (p = .498), but the conditional effects for those below 1 SD (p = .013) and within 1 SD from the mean (p =.005) again were found to be significant, but not for those 1 SD above the mean (p = .142).

Discussion

Findings of this study illustrate that when configuring smartphone apps, default effects work when providing privacy premium features with respect to autonomy (H1) and with respect to features supporting relatedness (H2) among adolescents. This study has extended the scope of existing research (Bischoff, 2008; Dogruel et al., 2017) in that it has shown that the default effect leads to about the same outcome with respect to two crucial characteristics of smartphone apps: privacy protection as well as fostering social relatedness (RQ1). The attractiveness of both types of features is rooted in self-determination theory, as they allow adolescents to satisfy innate needs, namely the need for autonomy and for relatedness, during their smartphone usage.

Overall, defaults led to an increase of around 40% in the preference for both types of app features. Adolescents seem to be equally persuaded by defaults for privacy as well as those for social relatedness. This is an encouraging result for privacy scholars. If someone wants to increase the selection of privacy protection features, it is helpful to know that these features do not lag behind other features that allow for more self-disclosure. The mechanisms to increase self-disclosure, as employed by social network sites in making disclosure a default and privacy an additional option (Acquisti et al., 2015; Stutzman et al., 2013), can just as easily be employed to create the inverse situation. When privacy protection becomes a default, adolescents tend to adhere to it in the same way as they adhere to the currently employed self-disclosure defaults. This study also found (RQ2) that sociodemographic as well as psychological variables only play a minor role and do not interfere with the observed default effect. This was also underlined by the fact that this study replicated the previous findings of Dogruel et al. (2017) with respect to defaults for privacy protection among adult American smartphone users. Additionally, the findings supported the assumption that girls, in general, are readier to protect their privacy than boys and that habitual app downloading may lead to an increase in the selection of social-relatedness features. However, default effects do not seem to be affected much by privacy concerns or literacy, as neither privacy concerns nor OPLIS correlated with the number of privacy premium features selected. Still, the fact that the sample is potentially too small to adequately identify these potential smaller sized effects must be acknowledged. Also, a meta-analysis on previous research has demonstrated that the relationship between privacy concerns and behavior is expected to be small (in the range of r = .1 to r = .2; Baruh et al., 2017). From a more exploratory perspective, the findings showed that age as an indicator for development could play a role. Not controlling for other factors, age correlated positively with a preference for privacy protection, and the exploratory analysis of RQ2 indicated that age could indeed be a moderator for default effects. Although the findings were inconclusive here, this study encourages further research to focus more on the moderating role of age. Consequently, it seems to be worthwhile to examine in future research if younger smartphone users are more susceptible to cognitive biases when selecting smartphone apps as compared to older adolescents. On a normative level, this could prove to be a useful finding for developing measures to counter common strategies used by platform providers and app developers that lead younger users to potentially disclose more information and protect their privacy less. As a consequence, policymakers should consider demanding more privacy protection defaults for social network sites (SNS) or apps, particularly those that cater to younger audiences.

Limitations

Before making such policy demands, we have to admit that the research has some methodological limitations. The research design was based on an existing study by Dogruel et al. (2017). This study replicated that design with respect to the app configuration task, which tested default effects for app selection. However, such a research design appears a bit artificial as was already noted by the authors. Apps usually come in a preset version, and users do not typically have the ability to change features according to their needs. Such an app configurator appears to be a viable instrument to test default effects, but the transferability of findings from such a test to a real-world setting remains limited. At the same time, the question arises about how adolescent participants might respond to such an abstract tool. On the one hand, changing an app in an online tool seems to be something that adolescents had few problems with since none were reported in the open-ended statements at the end of the questionnaire. Adolescents are used to such configuration tasks on SNS or computer games, and this paper focused on apps that were deemed to be enjoyed by adolescents. On the other hand, paying for an app is a concept that adolescents are potentially unfamiliar with in their real lives. The original study relied on a monetary valuation of app features because it made sense to suggest that premium features, such as better privacy protection, would come at a price. However, such costs being assigned a monetary value is still a rather abstract concept for most adolescents.

Another methodological flaw relates to the sample. This study relied on adolescents who voluntarily participated in the survey. Although the study had sufficient participants to test for medium-sized effects, a bigger sample would have provided enough data to also test for smaller sized effects, such as the relationship between privacy concerns and privacy behavior (Baruh et al., 2017). In the field, this study had a rather high dropout rate of more than 100 adolescents who opened the link to the survey but did not take part in it or provided incomplete data. As a likely result, the sample was biased towards better educated, “Gymnasium” students and females. Both groups are known to be more concerned about privacy (Madden et al., 2013). This bias was likely a consequence of the study’s sampling process, which in the end relied on self-selection on the part of the students. As often is found in comparable research, the study was unable to capture the responses from adolescents who do not care very much about their privacy. It would be interesting to see how the study would have played out with a lower educated sample that included more boys. That being said, these lower educated, less interested users are always more difficult to recruit.

Another limitation is more theoretical in nature. For answering RQ2, this study employed a rather exploratory approach by adding and testing several variables. These were variables that, to the researchers’ understanding, had been identified as influential for either privacy or for social-relatedness in apps. However, the researchers have to accept the criticism that such a selection is rather eclectic and not theory-driven. Research may continue to use this approach to further develop viable theoretical conceptualizations. For instance, as Valkenburg and Peter (2013) demonstrated on a theoretical level, children and adolescents have different susceptibilities towards media effects, and research needs to account for the complex interrelationships that might explain these differential susceptibilities.

Conclusion

Coming to a final conclusion, this research demonstrates that adolescents might be affected by cognitive biases when selecting features for smartphone apps. Defaults can work to increase their use of privacy protection features as well as to lead them to choose more features that enable them to disclose more information about themselves in exchange for social relatedness. Default effects are stable, but the question remains about the extent to which younger smartphone users are affected by them more so than older users.

Footnotes

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

ORCID iD

Sven Joeckel

Author biographies

Leyla Dogruel (PhD, Freie Universität Berlin) is junior-professor for Media Systems and Media Performance at Johannes Gutenberg-Universität Mainz, Germany. She was a visiting researcher at Ludwig-Maximilians-Universität München, Germany and at Wee Kim Wee School of Communication, Nanyang Technological University (NTU), Singapore. Her areas of research include media innovation as well as contexts of online privacy and algorithmic communication.

Sven Joeckel (PhD, TU Ilmenau) is professor for Communication with a Focus on Children, Adolescents, and the Media at the University of Erfurt, Germany. He has studied and worked at various German and international universities, including Dublin City University, Ireland and Wee Kim Wee School of Communication, Nanyang Technological University (NTU), Singapore. His research focuses on the use and effects of digital media, particularly on effects of mobile privacy regulations as well as moral reactions in and to digital media.

References

Acquisti

Brandimarte

Loewenstein

(2015). Privacy and human behavior in the age of information. Science, 347(6221), 509–514. doi:10.1126/science.aaa1465

Allen

K. A.

Ryan

Gray

D. L.

McInerney

D. M.

Waters

(2014). Social media use and social connectedness in adolescents: The positives and the potential pitfalls. The Australian Educational and Developmental Psychologist, 31(1), 18–31. doi:10.1017/edp.2014.2

Barnes

S. B.

(2006). A privacy paradox: Social networking in the United States. First Monday, 11(9). Retrieved from http://firstmonday.org/ojs/index.php/fm/article/view/1394/1312

Baruh

Secinti

Cemalcilar

(2017). Online privacy concerns and privacy management: A meta-analytical review. Journal of Communication, 67(1), 26–53. doi:10.1111/jcom.12276

Bischoff

(2008). Endowment effect theory, prediction bias and publicly provided goods: An experimental study. Environmental and Resource Economics, 39(3), 283–296. doi:10.1007/s10640-007-9126-3

boyd

(2014). It’s complicated: The social lives of networked teens. New Haven, CT: Yale University Press.

Brown

D. H.

Pecora

(2014). Online data privacy as a children’s media right: Toward global policy principles. Journal of Children and Media, 8(2), 201–207. doi:10.1080/17482798.2014.893756

Burgoon

J. K.

(1982). Privacy and communication. Annals of the International Communication Association, 6(1), 206–249. doi:10.1080/23808985.1982.11678499

Deci

E. L.

Ryan

R. M.

(2000). The “what” and “why” of goal pursuits: Human needs and the self-determination of behavior. Psychological Inquiry, 11(4), 227–268. doi:10.1207/S15327965PLI1104_01

10.

Deci

E. L.

Ryan

R. M.

(2012). Self-determination theory. In van Lange

P. A. M.

Kruglanski

A. W.

Higgins

E. T.

(Eds.), Handbook of theories of social psychology (pp. 416–437). Los Angeles, CA: SAGE.

11.

Derlega

V. J.

Metts

Petronio

Margulis

S. T.

(1993). Self-disclosure. Newbury Park, CA: SAGE.

12.

Dienlin

Metzger

M. J.

(2016). An extended privacy calculus model for SNSs: Analyzing self-disclosure and self-withdrawal in a representative U.S. sample. Journal of Computer-Mediated Communication, 21(5), 368–383. doi:10.1111/jcc4.12163

13.

Dienlin

Trepte

(2015). Is the privacy paradox a relic of the past? An in-depth analysis of privacy attitudes and privacy behaviors. European Journal of Social Psychology, 45(3), 285–297. doi:10.1002/ejsp.2049

14.

Dinev

Hart

(2006). An extended privacy calculus model for e-commerce transactions. Information Systems Research, 17(1), 61–80. doi:10.1287/isre.1060.0080

15.

Dogruel

Joeckel

Bowman

N. D.

(2015). Choosing the right app: An exploratory perspective on heuristic decision processes for smartphone app selection. Mobile Media & Communication, 3(1), 125–144. doi:10.1177/2050157914557509

16.

Dogruel

Joeckel

Vitak

(2017). The valuation of privacy premium features for smartphone apps: The influence of defaults and expert recommendations. Computers in Human Behavior, 77, 230–239. doi:10.1016/j.chb.2017.08.035

17.

Egelman

Felt

A. P.

Wagner

(2013). Choice architecture and smartphone privacy: There’s a price for that. In Böhme

(Ed.), The economics of information security and privacy (pp. 211–236). Berlin, Germany: Springer.

18.

Feierabend

Plankenhorn

Rathgeb

(2017). JIM-studie 2017. Jugend, information, (multi-)media [JIM-study 2017. Youth, information, (multi-)media]. Retrieved from https://www.mpfs.de/fileadmin/files/Studien/JIM/2017/JIM_2017.pdf

19.

Feng

Xie

(2014). Teens’ concern for privacy when using social networking sites: An analysis of socialization agents and relationships with privacy-protecting behaviors. Computers in Human Behavior, 33, 153–162. doi:10.1016/j.chb.2014.01.009

20.

Hayes

(2013). Introduction to mediation, moderation, and conditional process analysis: A regression based approach. New York, NY: Guilford Press.

21.

Johnson

E. J.

Bellman

Lohse

G. L.

(2002). Defaults, framing and privacy: Why opting in–opting out. Marketing Letters, 13(1), 5–15. doi:10.1023/A:1015044207315

22.

Jung

Kim

Chan-Olmsted

(2014). Measuring usage concentration of smartphone applications: Selective repertoire in a marketplace of choices. Mobile Media & Communication, 2(3), 352–368. doi:10.1177/2050157914542172

23.

Kahneman

Knetsch

J. L.

Thaler

R. H.

(1991). Anomalies: The endowment effect, loss aversion, and status quo bias. Journal of Economic Perspectives, 5(1), 193–206. doi:10.1257/jep.5.1.193

24.

Knijnenburg

Raybourn

Cherry

Wilkinson

Sivakumar

Sloan

(2017). Death to the privacy calculus? SSRN Electronic Journal. doi:10.2139/ssrn.2923806

25.

LaRose

. (2010). The problem of media habits. Communication Theory, 20(2), 194–222. doi:10.1111/j.1468-2885.2010.01360.x

26.

LaRose

Eastin

(2004). A social cognitive explanation of Internet uses and gratifications: Toward a new theory of media attendance. Journal of Broadcasting and Electronic Media, 48(3), 358–377.

27.

Laufer

R. S.

Wolfe

(1977). Privacy as a concept and a social issue: A multidimensional developmental theory. Journal of Social Issues, 33(3), 22–42. doi:10.1111/j.1540-4560.1977.tb01880.x

28.

Liu

Wei

(2014). Maintaining social connectedness in a fast-changing world: Examining the effects of mobile phone uses on loneliness among teens in Tibet. Mobile Media & Communication, 2(3), 318–334. doi:10.1177/2050157914535390

29.

Madden

Lenhart

Cortesi

Gasser

(2013). Teens and mobile apps privacy. Retrieved from http://www.pewinternet.org/reports/2013/teens-and-mobile-apps-privacy.aspx

30.

Margulis

S. T.

(2011). Three theories of privacy: An overview. In Trepte

Reinecke

(Eds.), Privacy online: Perspectives on privacy and self-disclosure in the social web (pp. 9–17). Berlin, Germany: Springer.

31.

Mascheroni

Ólafsson

(2014). Net Children Go Mobile: Risks and opportunities. Second edition. Retrieved from file:///C:/Users/MAC/Downloads/DEF_NCGM_SecondEdition_Report.pdf

32.

Masur

P. K.

Teutsch

Trepte

(2017). Entwicklung und Validierung der Online-Privatheitskompetenzskala (OPLIS) [Development and validation of the Online Privacy Competence Scale (OPLIS)]. Diagnostica, 63(4), 256–268. doi:10.1026/0012-1924/a000179

33.

Peter

Valkenburg

P. M.

(2011). Adolescents’ online privacy: Toward a developmental perspective. In Trepte

Reinecke

(Eds.), Privacy online: Perspectives on privacy and self-disclosure in the social web (pp. 221–234). Berlin, Germany: Springer.

34.

Petronio

(2002). Boundaries of privacy: Dialectics of disclosure (SUNY Series in Communication Studies). Albany: State University of New York Press.

35.

Pichert

Katsikopoulos

K. V.

(2008). Green defaults: Information presentation and pro-environmental behaviour. Journal of Environmental Psychology, 28(1), 63–73. doi:10.1016/j.jenvp.2007.09.004

36.

Reichheld

(2003). The one number you need to grow. Harvard Business Review, 81(12), 46–54.

37.

Rozendaal

Buijzen

Valkenburg

P. M.

(2009). Do children’s cognitive advertising defenses reduce their desire for advertised products? Communications, 34, 287–303.

38.

Shin

Kang

(2016). Adolescents’ privacy concerns and information disclosure online: The role of parents and the Internet. Computers in Human Behavior, 54, 114–123. doi:10.1016/j.chb.2015.07.062

39.

Stutzman

Gross

Acquisti

(2013). Silent listeners: The evolution of privacy and disclosure on Facebook. Journal of Privacy and Confidentiality, 4(2). doi:10.29012/jpc.v4i2.620

40.

Taddicken

(2014). The “privacy paradox” in the social web: The impact of privacy concerns, individual characteristics, and the perceived social relevance on different forms of self-disclosure. Journal of Computer-Mediated Communication, 19(2), 248–273. doi:10.1111/jcc4.12052

41.

Trepte

Reinecke

(2011). The social web as a shelter for privacy and authentic living. In Trepte

Reinecke

(Eds.), Privacy online: Perspectives on privacy and self-disclosure in the social web (pp. 61–74). Berlin, Germany: Springer.

42.

Trepte

Teutsch

Masur

P. K.

Eicher

Fischer

Hennhöfer

Lind

(2015). Do people know about privacy and data protection strategies? Towards the “Online Privacy Literacy Scale” (OPLIS). In Gutwirth

Leenes

de Hert

(Eds.), Law, Governance and Technology Series Vol. 20: Reforming European data protection law (pp. 333–366). Dordrecht, the Netherlands: Springer.

43.

Valkenburg

P. M.

Peter

(2013). The differential susceptibility to media effects model. Journal of Communication, 63(2), 221–243. doi:10.1111/jcom.12024

44.

Van Boven

Dunning

Loewenstein

. (2000). Egocentric empathy gaps between owners and buyers: Misperceptions of the endowment effect. Journal of Personality and Social Psychology, 79(1), 66–76.

45.

Van Gool

van Ouytsel

Ponnet

Walrave

. (2015). To share or not to share? Adolescents’ self-disclosure about peer relationships on Facebook: An application of the prototype willingness model. Computers in Human Behavior, 44, 230–239. doi:10.1016/j.chb.2014.11.036

46.

Vorderer

Krömer

Schneider

F. M.

(2016). Permanently online – Permanently connected: Explorations into university students’ use of social media and mobile smart devices. Computers in Human Behavior, 63, 694–703. doi:10.1016/j.chb.2016.05.085

47.

Walrave

Vanwesenbeeck

Heirman

(2012). Connecting and protecting? Comparing predictors of self-disclosure and privacy settings use between adolescents and adults. Cyberpsychology: Journal of Psychosocial Research on Cyberspace, 6(1). doi:10.5817/CP2012-1-3

48.

Westin

A. F.

(1967). Privacy and freedom. New York, NY: Atheneum.

49.

Rosson

M. B.

Gupta

Carroll

J. M.

(2012). Measuring mobile users’ concerns for information privacy. Proceedings of the 33rd Annual International Conference on Information Systems of the Association of Information Systems (AIS). Retrieved from http://faculty.ist.psu.edu/xu/papers/Xu_etal_ICIS_2012a.pdf

50.

Youn

(2009). Determinants of online privacy concern and its influence on privacy protection behaviors among young adolescents. Journal of Consumer Affairs, 43(3), 389–418. doi:10.1111/j.1745-6606.2009.01146.x