Comparing the Effect of Privacy and Non-Privacy Social Media Photo

Tools on Factors of Privacy Concern

Vanessa Bracamonte

, Sebastian Pape

and Sascha Loebner

KDDI Research, Inc., Saitama, Japan

Goethe University Frankfurt, Frankfurt am Main, Germany

Keywords:

Privacy Tools, Perceived Value of Information, Affect, Social Presence, Trust, Privacy Concern, User Study.

Abstract:

Research into privacy tools for social media content has found that although there is a positive attitude towards

these tools, there are also privacy concerns related to the user information involved. This privacy concern

towards privacy tools can be higher than for non-privacy tools of a similar type, but the reason for this differ-

ence is not clear. To address this, we conducted an online experiment to compare the effect of a privacy and

a non-privacy tool on antecedent factors of privacy concern (Perceived value of personal information, Social

presence, Affect and Trust) which are hypothesized to be affected by the different purpose of the tools. The

results show that participants had higher affect towards the privacy tool compared to the non-privacy tool. On

the other hand, the results also show that participants in the privacy tool group had a higher level of perception

of value of their personal information, that is, that the information provided to and inferred by the tool is valu-

able. Finally, both factors mediated a signiﬁcant but opposing effect of the type of tool on privacy concern.

1 INTRODUCTION

Automated analysis of images has been proposed as

a way of protecting peoples’ privacy in a social me-

dia context (Korayem et al., 2016). In general terms,

these proposals work by analyzing the content of

users’ photos to detect whether the content reveals

private or sensitive information and potentially trans-

forming that content to anonymize it (Ilia et al., 2015;

Li et al., 2019; Hasan et al., 2020). Research has

identiﬁed that users worry about their data being col-

lected, sold, shared or misused, and about having their

privacy intruded upon by these privacy tools (Braca-

monte et al., 2021; Bracamonte et al., 2022). Al-

though in essence this type of privacy tool would

not be very different from a tool that analyzes users’

content for the purposes of enjoyment, research has

found that privacy concerns towards privacy tools can

be higher than towards tools that would also analyze

user data, but which have a non-privacy related pur-

pose (Bracamonte et al., 2022). This indicates that

the priming of the privacy tool has an effect on privacy

concerns, but it is unclear through which mechanisms

these effects are occurring.

We hypothesize that a number of factors may have

a different effect on privacy concern towards privacy

tools compared to non-privacy tools, and conduct an

online survey-based experiment comparing the per-

ception of two tools with similar characteristics but a

different purpose: a tool for privacy vs. a non-privacy

tool for enjoyment. In this paper, we evaluate possi-

ble constructs through which the positive and negative

perceptions may be inﬂuencing privacy concern: Per-

ceived value of personal information, Affect and Trust

towards the tool, and Social presence.

2 RELATED WORK

Research on perception towards privacy tools that

process personal data has identiﬁed that users have

privacy concerns towards these types of tools. (Bra-

camonte et al., 2022) conducted a study comparing

privacy and non-privacy tools for text and photos, and

found that these concerns can be higher than for sim-

ilar tools with a non-privacy related purpose. They

also found that similar themes are reported when ex-

plaining the reasons for privacy concern towards both

types of tools.

However, existing research has not evaluated

through which mechanism do privacy tools inﬂuence

privacy concerns differently than non-privacy tools.

Nevertheless, ﬁndings from user evaluation studies of

Bracamonte, V., Pape, S. and Loebner, S.

Comparing the Effect of Privacy and Non-Privacy Social Media Photo Tools on Factors of Privacy Concern.

DOI: 10.5220/0011784900003405

In Proceedings of the 9th International Conference on Information Systems Security and Privacy (ICISSP 2023), pages 669-676

ISBN: 978-989-758-624-8; ISSN: 2184-4356

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

669

privacy tools suggest some possibilities of variables

which may explain this difference.

2.1 Perceived Value of Information

Privacy tools can cause some users to become con-

cerned about how the tool itself may obtain value

from the data they collect. (Schaub et al., 2016)

found that users have reported concerns that privacy

tools could potentially proﬁt from their personal in-

formation (Bracamonte et al., 2022) similarly found

that some users were concerned that a privacy tool

could proﬁt from their information through advertis-

ing, sharing or selling of the data.

When users become aware that their personal in-

formation is valuable to third parties, this can cause

them to place a higher value on that information

(Spiekermann et al., 2012). (Danezis et al., 2005)

found that students increased their bids for the dis-

closure of their location information after learning

that there was a possible commercial interest in that

data. (Staiano et al., 2014) found that aggregated per-

sonal identiﬁable information, collected over a period

of time, was valued higher than individual points of

data , and it was hypothesized that the higher valua-

tion was due to the participants realizing that this data

revealed more information about their life.

A privacy tool such as the one described in this pa-

per would work by detecting whether the information

contained in the users’ photos may be privacy sensi-

tive. Therefore, the tool itself assigns some meaning

to the content of the photo, which the users’ may not

have been aware of in advance. In doing so, the tool

may be indicating that the information in the photo

is something valuable to be protected. (Bracamonte

et al., 2022) found that users appear to be concerned

not only about the data collection aspect, but also

what the privacy tool could gain from the data. If

the photos are something to be protected, rather than

“normal” photos, then it is possible that users would

consider that data more valuable, to themselves or to

others who would proﬁt from it. This could result in

an increased perception of the value of personal in-

formation, in comparison with tools that do not have

a privacy-enhancing purpose. The privacy tool’s in-

ference or prediction of the users’ content as (poten-

tially) privacy sensitive may increase the perception

of how valuable the content is, since the tools indi-

cates it needs to be protected.

2.2 Social Presence

Social presence has been investigated as an inﬂuence

on the positive perception of technology. (Gefen and

Straub, 2004) indicate that social presence can have

effect of perception in situations of interaction with

technology where there is typically no direct interac-

tion with a person. For example, the sense that others

are present helps increase factors such as perceived

trust (Gefen and Straub, 2004; Kim et al., 2013; Has-

sanein and Head, 2005) and beneﬁt (Kim et al., 2013).

The effect of social presence is present in different

types of contexts of technology mediation, from web-

sites (Gefen and Straub, 2004) to automation tech-

nology such as robots (Kim et al., 2013), to varying

degrees. Much of the research has focused on the

effect of human images, but research by (Hassanein

and Head, 2005) has shown that interfaces that evoke

emotions such as fun or interest can also increase the

perception of social presence.

An utilitarian tool such a tool for privacy may not

have as much social presence, in the sense of posi-

tive human connection, in comparison with a tool for

enjoyment. In addition, (Oh et al., 2018) report that

social presence is not always correlated with positive

effects, and that its effects may depend on the context,

such as when a person is vulnerable. It is likely that

users would imagine different contexts for the use of

privacy vs. non-privacy tools. In a privacy context,

users may feel vulnerable and social presence could

represent a risk rather than a beneﬁt.

2.3 Affect and Trust

It is also important to consider that not only negative

perceptions could affect privacy concern. A privacy

tool such as the one in the current study only dif-

fers from a non-privacy tool in its purpose, and be-

cause the purpose is privacy protection or enhance-

ment, which beneﬁts the user, this may result in posi-

tive perception. (Dinev et al., 2015) indicate that pri-

vacy concerns are not only inﬂuenced by analytical

processes, but are also affected by dispositional atti-

tudes. Extraneous inﬂuences have an effect on privacy

concerns, and positive or negative feelings and heuris-

tics are also used used in order to decide and make pri-

vacy judgments about a particular technology. Users

have positive feelings towards privacy tools due to

their purpose of protection (Schaub et al., 2016; Bra-

camonte et al., 2021). Positive feelings, inﬂuenced by

the beneﬁcial purpose of this type of tools, could also

work to reduce concerns. (Schaub et al., 2016) in-

dicated that users had a feeling of protection by the

privacy tool, and that this mitigated concerns related

ICISSP 2023 - 9th International Conference on Information Systems Security and Privacy

670

to tracking. (Bracamonte et al., 2022) found that the

purpose of the privacy tool was mentioned as one of

the reasons for having a lower level of privacy con-

cern, whereas there was no such reason for the non-

privacy tool. The privacy tool, which purpose is to

protect users, could increase positive feelings in users.

Positive feelings, or Affect, can be deﬁned as

thought of as automatic responses associated with a

stimulus (Slovic et al., 2002). If we compare a privacy

and non-privacy tool, we can take the privacy purpose

of the former as a stimulus that primes users for a

type of reaction or response. Therefore, in their ini-

tial interaction with a privacy tool, positive perception

of the privacy tool brought by feelings of protection

may increase affect towards it, compared to a non-

privacy tool. Furthermore, (Finucane et al., 2000)

and (Slovic et al., 2002) indicate that overall affective

evaluation of an object can inﬂuence the risk judg-

ment related to that object. For example, (Kehr et al.,

2015) found that positive feelings towards a technol-

ogy medium can inﬂuence people’s privacy assess-

ment and can result in underestimating information

disclosure risks. Another factor which can automat-

ically affected by stimulus is trust (Lindgaard et al.,

2006). (Gefen and Straub, 2004) indicate that trust

plays an important role in the perception of technol-

ogy, and it is considered important for privacy tools

as well (Balebako et al., 2013; Coopamootoo, 2020;

Harborth et al., 2020).

3 METHODOLOGY

3.1 Research Questions

The objective of the paper is to evaluate the mech-

anism through which a privacy tool affects privacy

concern differently from a non-privacy tool, focusing

speciﬁcally on tools for the analysis and transforma-

tion of social media photos. We propose a number

of factors to test, and evaluate the following research

questions:

R1: Are the levels of Perceived value of informa-

tion, Social presence, Affect and Trust different when

viewing a privacy tool compared to viewing a non-

privacy tool?

R2: Does the privacy tool have a different ef-

fect (compared to the non-privacy tool) on Privacy

concern through those variables? That is, do Per-

ceived value of information, Social presence, Affect

and Trust mediate the effect of the type of tool on Pri-

vacy concern?

3.2 Experiment Design

In order to answer the research questions, we de-

signed an experiment which consisted of a task for

participants to view, read a description and give their

opinion about an hypothetical app that would be used

to transform photos for uploading on social media.

We manipulated the type of tool that the participants

viewed: a privacy tool or a non-privacy tool for so-

cial media photos. The objective of this study was to

evaluate differences in perception that resulted from

the manipulation (priming), therefore, the privacy tool

was explicitly described as such. Participants viewed

the description and a mockup of only one type of app

(between-subjects design). After reading about the

app, the participants answered a questionnaire.

3.3 Task

We described to participants an hypothetical free,

third-party app for social media photos. For the pri-

vacy tool, the purpose was described as protecting pri-

vacy; for the non-privacy tool, the purpose was de-

scribed as enhancing the content for fun. The app

would hypothetically work by analyzing and detect-

ing the content in the users’ photos. We described

the type of information the app would detect from

the photos: private information for the privacy tool,

and information that could be enhanced with stickers

for the non-privacy tool. We then presented a non-

interactive mockup of the app interface which showed

how it would work. The mockups for each group had

the same general design, and only differed in their

message (“Privacy alert!” vs “Enhance it!”) and

the transformation performed on the photo (privacy-

enhancing vs non-privacy-enhancing).

After the mockup, we showed ﬁve additional

photo examples to the participants. The photos for

the examples were obtained from the COCO dataset

(Lin et al., 2014).

3.4 Measurement Items

We included items adapted from previous research to

measure Social presence (Gefen and Straub, 2004)

and Affect (Kehr et al., 2015). The items for mea-

suring Perceived value of information were developed

based on the single-item measurement from (Spieker-

mann et al., 2012), adapted to refer to personal infor-

mation provided by the user and personal information

generated by the tool. Privacy concern was measured

with the Mobile Users’ Information Privacy Concerns

(MUIPC) scale, which is comprised of the dimen-

sions of Perceived surveillance, Perceived intrusion

Comparing the Effect of Privacy and Non-Privacy Social Media Photo Tools on Factors of Privacy Concern

671

and Secondary use of personal information (Xu et al.,

2012). Trust was measured with 3 items adapted

from (Jarvenpaa et al., 1999). All questions had

a 7-point response scale, from Strongly disagree to

Strongly agree, with the exception of the Affect items.

To validate that the samples in each group were

comparable, we included questions on Prior pri-

vacy experience (Smith et al., 1996), Disposition to

value privacy (Xu et al., 2011), Information sensitiv-

ity (Dinev et al., 2013), as well as questions on the

participants’ age, gender (as an open text box (Spiel

et al., 2019)) and frequency of social media posting,

in general and for photos in particular. The question-

naire also included open-ended attention check ques-

tions, where we asked participants to answer brieﬂy

about the app described.

3.5 Ethical Considerations

This study was exempt from review according to our

institution’s criteria for research of this type. Never-

theless, we provided a notice to inform potential par-

ticipants about the characteristics of the study. The

notice included a description of the purpose of the

survey, the approximate time to ﬁnish it and the task

that participants were expected to do (read a descrip-

tion and answer questions). The notice also explained

that the survey included attention questions, but that

we would not reject the participants’ answers based

only on these questions. However, we clariﬁed that

we would reject duplicated answers or answers unre-

lated to the question asked.

We indicated that the survey was completely vol-

untary and that participants were free to decline to

participate, that we would not collect identifying in-

formation such as name, email or IP address, and that

the results would be used for academic purposes only.

We also indicated that the survey was limited to adults

who lived in the United States. Finally, we provided

the principal researcher’s name and email address in

case of any questions about the study. Participants

were asked to access the link to the survey itself if

they accepted to participate.

3.6 Participant Recruitment

We recruited participants by posting a task to answer

a survey on Amazon Mechanical Turk. We set the

qualiﬁcations for participation for workers from the

USA, who had a 99% rate of acceptance rate for their

tasks and who had worked on at least 5000 tasks. We

set the participant reward at US$2.5.

The survey ran on February 8-9, 2022. We ob-

tained 400 responses in total. We reviewed the an-

swers to all open-ended questions to identify cases

with multiple nonsensical answers or answers which

were completely unrelated to the questions. We iden-

tiﬁed 20 such responses, and rejected them. The

rest of the participants were rewarded, at a rate of

US$12.5/hour. The median response time was 12

minutes.

4 RESULTS

In this section, we describe the analysis method (PLS-

SEM) we used, then report the ﬁnal sample character-

istics, the measurement model’s reliability and valid-

ity analysis results, and ﬁnally the results of the struc-

tural model analysis.

To evaluate the research questions of this study,

we used the partial least squares structural equation

modeling (PLS-SEM) method. In summary, the PLS-

SEM method consists of a series of steps from check-

ing the reliability of the measurement model (rela-

tionship between the measurement items and its cor-

responding construct), the reliability of the struc-

tural model (relationships between the constructs)

and ﬁnally obtaining the path coefﬁcients (direct and

mediated) i.e. the relationships between the con-

structs (Hair et al., 2019).

We conducted the PLS-SEM analyses using the R

seminr package (Ray et al., 2022).

4.1 Sample Characteristics

The sample consisted of 380 cases. We ﬁrst identi-

ﬁed 22 multivariate outliers, using the Mahalanobis

distance (alpha = 0.001), and removed them. We

reviewed the answers to the attention questions, but

found none that could be considered incorrect. The

ﬁnal sample for analysis consisted of 358 participant

responses, 179 in each group. The sample size ob-

tained was over the minimum for ﬁnding path coefﬁ-

cients of 0.11 - 0.2 with a 1% signiﬁcance level and a

power of 80% with the PLS-SEM method (Hair et al.,

2021), according to guidelines based on the inverse

square root method for minimum sample size estima-

tion (Kock and Hadaya, 2018). The age mean was

41 years-old for both groups. The gender distribution

was 53% female / 48% male participants in the pri-

vacy tool group, and 56% female and 44% male in

the non-privacy tool group.

Mann-Whitney U tests indicated no signiﬁcant

differences in age (W = 16124, p-value = 0.92) fre-

quency of social media posting, in general (W =

17594, p-value = 0.10) or photos in particular (W =

17446, p-value = 0.14), Prior privacy experience (W

ICISSP 2023 - 9th International Conference on Information Systems Security and Privacy

672

= 14904, p-value = 0.25), Disposition to value privacy

(W = 14960, p-value = 0.28) or Information sensitiv-

ity (W = 15230, p-value = 0.42) between groups.

4.2 Measurement Model

We ﬁrst evaluated the reliability and validity of the

measurement model. The PLS-SEM method can

be used to evaluate multiple relationships simultane-

ously, so in addition to the relationships related to the

research questions, we also control for the effects of

constructs which have been validated in previous re-

search, such as the inﬂuence on Social presence on

Trust (Gefen and Straub, 2003), but we do not focus

on those in this paper.

An initial evaluation indicated that although all

other reliability indicators were satisfactory, the con-

structs of Perceived surveillance, Intrusion and Sec-

ondary use of personal information had low discrimi-

nant validity. These constructs are dimensions of Pri-

vacy concern (Xu et al., 2012), and therefore concep-

tually similar. Therefore, we modeled Privacy con-

cern as a higher-order construct and report the re-

sults of the evaluation of the higher-order measure-

ment model. We examined indicator reliability by

inspecting the values obtained by squaring the item

loadings, which are the correlation weights between

the construct and its indicators (measurement items).

All items had an indicator value (squared loading)

over the threshold of .708 (Hair et al., 2019).

To evaluate internal consistency reliability, which

is the association between indicators of the same con-

struct, we examined the Cronbach’s alpha and the

rhoA reliability coefﬁcient (Dijkstra and Henseler,

2015). For all constructs, Cronbach’s alpha values

ranged from 0.932 - 0.973, and rhoA values ranged

from 0.932 - 0.973, which is higher than the satisfac-

tory minimum of 0.7 (Hair et al., 2021). Although re-

liability coefﬁcient values were higher than the ideal

upper limit of 0.9, this is likely due to the use of es-

tablished scales.

Convergent validity, which is how much the con-

struct converges to explain indicator variance, was ex-

amined using the average variance extracted (AVE).

The AVE for all constructs had a value above the min-

imum level of 0.5 (Hair et al., 2019), on a range from

0.856 - 0.949. Finally, we examined the discriminant

validity, which is how much a construct is distinct

from other constructs, using the heterotrait–monotrait

(HTMT) ratio of correlations criterion. All values

were signiﬁcantly lower than the threshold value of

0.9 (Hair et al., 2021).

4.3 Structural Model

After validating the measurement model, we assessed

the structural model. We ﬁrst examined whether there

were collinearity issues (too high correlation between

constructs) in the structural model, by calculating the

variance inﬂation factor (VIF) values for the con-

structs. All VIF values were below the minimum

recommended of 3 (Hair et al., 2021), with values

ranging from 1.01 - 2.11. We then examined the R-

squared values, which measure the variance in a con-

struct explained by the predictors; values of 0.25 are

considered weak (Hair et al., 2019). Social presence

and Perceived value of information were the only con-

structs with a value lower than 0.25, but this is ex-

pected since the type of tool is their only predictor in

the model. All other values were above the moderate

threshold of 0.5 (Hair et al., 2019).

The structural model included the experiment

groups as a dichotomous variable (1 for the pri-

vacy tool and 0 for the non-privacy tool). We con-

ducted a bootstrapping procedure with 10,000 sam-

ples (Sarstedt et al., 2016) to calculate the path signif-

icance. Statistical signiﬁcance criteria in this case is

determined by the bootstrapped standardized t statis-

tic (Hair et al., 2021): above 3.291 corresponds to sig-

niﬁcant at 0.1% probability of error (alpha = 0.001) ;

above 2.576, to signiﬁcant at 1% (alpha = 0.01), and

above 1.96, to signiﬁcant at 5% (alpha = 0.05) (two-

tailed). These are represented in the tables with * for

signiﬁcant at 5%, ** for 1%, and *** for 0.1%.

Figure 1 shows the representation of those results.

The results, speciﬁcally the direct paths from the vari-

able representing the tool, answer the ﬁrst research

question (R1) of whether the level of the factors is

different for the privacy tool compared to the non-

privacy tool group. The results show that Perceived

value of information and Affect were signiﬁcantly

higher for the privacy tool. On the other hand, Social

presence, Trust and also Privacy concern itself were

not signiﬁcantly different between the two groups.

Although they are not the focus of this study, we ob-

serve from the results that other direct relationships

between constructs were signiﬁcantly different. How-

ever, one interesting result is that Social presence in-

creases Privacy concern, but also increases Affect and

Trust.

We then examined the mediated effects on Pri-

vacy concern (Table 1) to answer the second research

question (R2). The results show that there was a sig-

niﬁcant indirect effect of the privacy tool on Privacy

concern (compared to the non-privacy tool) through

Perceived value of information, which resulted in an

increase of Privacy concern. There was also a signif-

Comparing the Effect of Privacy and Non-Privacy Social Media Photo Tools on Factors of Privacy Concern

673

𝛽

= 0.216***

95%CI [0.137,0.296]

𝛽 = 0.177***

95%CI [0.074,0.275]

𝛽

= -0.07

95%CI [-0.176,0.035]

𝛽 = 0.004

95%CI [-0.075,0.083]

𝛽

= 0.026

95%CI [-0.051,0.103]

𝛽 = 0.488***

95%CI [0.372,0.596]

𝛽

= -0.216***

95%CI [-0.312,-0.117]

𝛽 = 0.336***

95%CI [0.255,0.421]

𝛽

= 0.602***

95%CI [0.526,0.673]

𝛽

= 0.23***

95%CI [0.114,0.348]

𝛽 = 0.19***

95%CI [0.097,0.278]

𝛽 = -0.596***

95%CI [-0.693,-0.491]

Figure 1: Results of the PLS-SEM analysis.

Table 1: Mediation effects results: standardized coefﬁcients and signiﬁcance.

X Mediator(s) Original

Est

Bootstrap

Mean

Bootstrap

T Stat. 95% CI Signif.

Tool Value 0.059 0.06 0.02 3.034 [ 0.023, 0.1 ] **

Tool SocialPresence −0.013 −0.013 0.011 −1.193 [ −0.037, 0.006]

Tool Affect −0.047 −0.047 0.014 −3.376 [ −0.077, −0.022] ***

Tool Trust −0.002 −0.003 0.024 −0.086 [ −0.051, 0.043]

icant mediated effect on Privacy concern through Af-

fect which resulted in a decrease of Privacy concern.

In line with the direct effect results, neither Trust nor

Social presence signiﬁcantly mediate the effect of the

privacy tool on Privacy concern.

5 DISCUSSION

The results of the analysis indicate that privacy con-

cern towards a privacy tool is affected through an in-

crease of the perception of value of the information

that the user gives the tool and the value of the infor-

mation the tool infers from the photos, and through

Affect and Trust towards the privacy tool, which is

higher in comparison to the non-privacy tool. How-

ever, these variables have opposing effects on privacy

concern. Affect decreased Privacy concern and acted

as a mediator of the positive effect of Social pres-

ence on Privacy concern. In the opposite direction,

an increased Perception of the value of personal in-

formation increased Privacy concern towards the tool,

and countered the previously described effect. In this

study, the result was that the overall effect the privacy

tool on privacy concern (compared to the non-privacy

tool) was not signiﬁcant (unlike the ﬁndings by (Bra-

camonte et al., 2022)), as these two opposing effects

canceled each other out.

We consider that the fact that users report a higher

a level of Perceived value of information represents

an interesting problem for privacy tools of the type

we focus on in this study. We can assume that the

provider of a such a tool would naturally want to em-

phasize the privacy protection aspect, and this may

increase a positive attitude towards the tool. On the

other hand, by emphasizing protection the providing

may also be emphasizing the value of the data that is

to be protected. Once that is established, users may be

able to more easily imagine what would be the nega-

tive consequences of losing control of that data. An-

other possibility is that the provider could emphasize

other aspects of the privacy tool besides the protection

itself, in order to increase affect towards it.

Finally, although the results showed that the pri-

vacy tool did not signiﬁcantly increase or decrease

perception of Social presence compared to the non-

privacy tools, Social presence directly increased Pri-

vacy concern and at the same time increased positive

perceptions (Affect and Trust). This might indicate

that a variable not included in this study may be me-

diating the effect of Social presence on Privacy Con-

cern. Future research should validate empirically val-

ICISSP 2023 - 9th International Conference on Information Systems Security and Privacy

674

idate these effects, since research has shown that the

effect of Social presence can depend on the context of

use of a technology (Phelan et al., 2016; Zhang and

Xu, 2016; Mozafari et al., 2021).

5.1 Limitations

First, we used a non-interactive app mockup for the

experiment. This decreases the realism of the situ-

ation for participants, who are not risking their pri-

vate information. Second, we only evaluated a lim-

ited set of variables that affect Privacy concern in this

experiment. The results show that these variables ex-

plained Privacy Concern with an R-squared = 0.538

(moderate). Nevertheless, we acknowledge that there

could be other variables which could also explain a

difference in perception of these two types of tools.

Third, we recruited Amazon Mechanical Turk work-

ers for the experiment. Although research has found

that these workers have a higher sensitivity to pri-

vacy issues (Kang et al., 2014), there is also evidence

that privacy-related related knowledge matches the

US population to some extent (Redmiles et al., 2019).

Nevertheless, the results might not generalize to other

populations.

6 CONCLUSIONS

Previous research has found evidence of a higher level

of privacy concern towards privacy tools compared to

similar tools with a non-privacy purpose. In this pa-

per, we explore the mechanisms that might explain

this difference in privacy concern. We found that pri-

vacy tools increased the Perception of value of per-

sonal information, and that this variable mediated an

increase of Privacy concern caused by the privacy

tool. We also found that the privacy tool increases Af-

fect compared to the non-privacy tool, and that Affect

along with Trust mediate a decrease of Privacy con-

cern. In our study, these effects appeared to cancel

each other out, resulting in an overall non-signiﬁcant

effect of the privacy tool on Privacy concern, com-

pared to the non-privacy tool. Finally, we found no

signiﬁcant differences in the level of Trust and Social

presence between the privacy and non-privacy tools,

although these two variables had an effect on Privacy

concern.

In future research, we plan to empirically exam-

ine how privacy tools affect the perception of value of

information, and how this perception affects privacy

concern in turn.

REFERENCES

Balebako, R., Jung, J., Lu, W., Cranor, L. F., and Nguyen,

C. (2013). ”Little Brothers Watching You”: Raising

awareness of data leaks on smartphones. In Proceed-

ings of the Ninth Symposium on Usable Privacy and

Security, SOUPS ’13, pages 1–11. Association for

Computing Machinery.

Bracamonte, V., Pape, S., and Loebner, S. (2022). “All apps

do this”: Comparing privacy concerns towards privacy

tools and non-privacy tools for social media content.

Proc. Priv. Enhancing Technol., 2022(3).

Bracamonte, V., Tesfay, W. B., and Kiyomoto, S. (2021).

Towards Exploring User Perception of a Privacy Sen-

sitive Information Detection Tool. In 7th International

Conference on Information Systems Security and Pri-

vacy.

Coopamootoo, K. P. (2020). Usage Patterns of Privacy-

Enhancing Technologies. In Proceedings of the 2020

ACM SIGSAC Conference on Computer and Commu-

nications Security, CCS ’20, pages 1371–1390. Asso-

ciation for Computing Machinery.

Danezis, G., Lewis, S., and Anderson, R. (2005). How

much is location privacy worth. In In Proceedings of

the Workshop on the Economics of Information Secu-

rity Series (WEIS.

Dijkstra, T. K. and Henseler, J. (2015). Consistent Par-

tial Least Squares Path Modeling. MIS Quarterly,

39(2):297–316.

Dinev, T., McConnell, A. R., and Smith, H. J. (2015).

Research Commentary—Informing Privacy Research

Through Information Systems, Psychology, and Be-

havioral Economics: Thinking Outside the “APCO”

Box. Information Systems Research, 26(4):639–655.

Dinev, T., Xu, H., Smith, J. H., and Hart, P. (2013). Informa-

tion privacy and correlates: An empirical attempt to

bridge and distinguish privacy-related concepts. Euro-

pean Journal of Information Systems, 22(3):295–316.

Finucane, M. L., Alhakami, A., Slovic, P., and Johnson,

S. M. (2000). The affect heuristic in judgments of

risks and beneﬁts. Journal of behavioral decision

making, 13(1):1–17.

Gefen, D. and Straub, D. (2003). Managing user trust in

B2C e-services. e-Service, 2(2):7–24.

Gefen, D. and Straub, D. (2004). Consumer trust in B2C

e-Commerce and the importance of social presence:

Experiments in e-Products and e-Services. Omega,

32:407–424.

Hair, J., Hult, G. T. M., Ringle, C., Sarstedt, M., Danks, N.,

and Ray, S. (2021). Partial Least Squares Structural

Equation Modeling (PLS-SEM) Using R: A Workbook.

Hair, J. F., Risher, J. J., Sarstedt, M., and Ringle, C. M.

(2019). When to use and how to report the results of

PLS-SEM. European business review.

Harborth, D., Pape, S., and Rannenberg, K. (2020). Ex-

plaining the technology use behavior of privacy-

enhancing technologies: The case of tor and Jon-

Donym. Proceedings on Privacy Enhancing Tech-

nologies (PoPETs), 2020(2):111–128.

Comparing the Effect of Privacy and Non-Privacy Social Media Photo Tools on Factors of Privacy Concern

675

Hasan, R., Crandall, D., Fritz, M., and Kapadia, A. (2020).

Automatically Detecting Bystanders in Photos to Re-

duce Privacy Risks. In 2020 IEEE Symposium on Se-

curity and Privacy (SP), pages 318–335.

Hassanein, K. and Head, M. (2005). The impact of infusing

social presence in the web interface: An investigation

across product types. International Journal of Elec-

tronic Commerce, 10(2):31–55.

Ilia, P., Polakis, I., Athanasopoulos, E., Maggi, F., and Ioan-

nidis, S. (2015). Face/off: Preventing privacy leak-

age from photos in social networks. In Proceedings of

the 22nd ACM SIGSAC Conference on Computer and

Communications Security, pages 781–792.

Jarvenpaa, S. L., Tractinsky, N., and Saarinen, L. (1999).

Consumer Trust in an Internet Store: A Cross-Cultural

Validation. Journal of Computer-Mediated Communi-

cation, 5(2):JCMC526.

Kang, R., Brown, S., Dabbish, L., and Kiesler, S. (2014).

Privacy Attitudes of Mechanical Turk Workers and the

U.S. Public. In 10th Symposium On Usable Privacy

and Security ({SOUPS} 2014), pages 37–49.

Kehr, F., Kowatsch, T., Wentzel, D., and Fleisch, E. (2015).

Blissfully ignorant: The effects of general privacy

concerns, general institutional trust, and affect in the

privacy calculus: Privacy calculus: Dispositions and

affect. Information Systems Journal, 25(6):607–635.

Kim, K. J., Park, E., and Sundar, S. S. (2013). Caregiving

role in human–robot interaction: A study of the medi-

ating effects of perceived beneﬁt and social presence.

Computers in Human Behavior, 29(4):1799–1806.

Kock, N. and Hadaya, P. (2018). Minimum sample size

estimation in PLS-SEM: The inverse square root and

gamma-exponential methods. Information systems

journal, 28(1):227–261.

Korayem, M., Templeman, R., Chen, D., Crandall, D., and

Kapadia, A. (2016). Enhancing Lifelogging Privacy

by Detecting Screens. In Proceedings of the 2016 CHI

Conference on Human Factors in Computing Systems,

CHI ’16, pages 4309–4314, New York, NY, USA. As-

sociation for Computing Machinery.

Li, F., Sun, Z., Li, A., Niu, B., Li, H., and Cao, G. (2019).

HideMe: Privacy-Preserving Photo Sharing on Social

Networks. In IEEE INFOCOM 2019 - IEEE Confer-

ence on Computer Communications, pages 154–162.

Lin, T.-Y., Maire, M., Belongie, S., Hays, J., Perona, P., Ra-

manan, D., Doll

ar, P., and Zitnick, C. L. (2014). Mi-

crosoft COCO: Common Objects in Context. In Fleet,

D., Pajdla, T., Schiele, B., and Tuytelaars, T., edi-

tors, Computer Vision – ECCV 2014, Lecture Notes

in Computer Science, pages 740–755. Springer Inter-

national Publishing.

Lindgaard, G., Fernandes, G., Dudek, C., and Brown, J.

(2006). Attention web designers: You have 50 mil-

liseconds to make a good ﬁrst impression! Behaviour

& information technology, 25(2):115–126.

Mozafari, N., Weiger, W., and Hammerschmidt, M. (2021).

That’s so Embarrassing! When Not to Design for So-

cial Presence in Human-Chatbot Interactions.

Oh, C. S., Bailenson, J. N., and Welch, G. F. (2018). A Sys-

tematic Review of Social Presence: Deﬁnition, An-

tecedents, and Implications. Frontiers in Robotics and

AI, 5.

Phelan, C., Lampe, C., and Resnick, P. (2016). It’s Creepy,

But it Doesn’t Bother Me. In Proceedings of the 2016

CHI Conference on Human Factors in Computing Sys-

tems, pages 5240–5251. Association for Computing

Machinery.

Ray, S., Danks, N. P., Valdez, A. C., Estrada, J. M. V., Uan-

horo, J., Nakayama, J., Koyan, L., Burbach, L., Bejar,

A. H. C., and Adler, S. (2022). Seminr: Building and

Estimating Structural Equation Models.

Redmiles, E. M., Kross, S., and Mazurek, M. L. (2019).

How well do my results generalize? comparing secu-

rity and privacy survey results from mturk, web, and

telephone samples. In 2019 IEEE Symposium on Se-

curity and Privacy (SP), pages 1326–1343.

Sarstedt, M., Hair, J. F., Ringle, C. M., Thiele, K. O., and

Gudergan, S. P. (2016). Estimation issues with PLS

and CBSEM: Where the bias lies! Journal of Business

Research, 69(10):3998–4010.

Schaub, F., Marella, A., Kalvani, P., Ur, B., Pan, C., Forney,

E., and Cranor, L. F. (2016). Watching Them Watch-

ing Me: Browser Extensions Impact on User Privacy

Awareness and Concern. In Proceedings 2016 Work-

shop on Usable Security. Internet Society.

Slovic, P., Finucane, M., Peters, E., and MacGregor, D. G.

(2002). Rational actors or rational fools: Implications

of the affect heuristic for behavioral economics. The

Journal of Socio-Economics, 31(4):329–342.

Smith, H. J., Milberg, S. J., and Burke, S. J. (1996).

Information Privacy: Measuring Individuals’ Con-

cerns about Organizational Practices. MIS Quarterly,

20(2):167–196.

Spiekermann, S., Korunovska, J., and Bauer, C. (2012).

Psychology of ownership and asset defense: Why

people value their personal information beyond pri-

vacy. Available at SSRN 2148886.

Spiel, K., Haimson, O. L., and Lottridge, D. (2019). How

to do better with gender on surveys: A guide for HCI

researchers. Interactions, 26(4):62–65.

Staiano, J., Oliver, N., Lepri, B., de Oliveira, R., Caraviello,

M., and Sebe, N. (2014). Money walks: A human-

centric study on the economics of personal mobile

data. In Proceedings of the 2014 ACM International

Joint Conference on Pervasive and Ubiquitous Com-

puting, pages 583–594. ACM.

Xu, H., Dinev, T., Smith, J., and Hart, P. (2011). Infor-

mation Privacy Concerns: Linking Individual Percep-

tions with Institutional Privacy Assurances. Journal

of the Association for Information Systems, 12(12).

Xu, H., Gupta, S., Rosson, M., and Carroll, J. (2012). Mea-

suring Mobile Users’ Concerns for Information Pri-

vacy. In ICIS.

Zhang, B. and Xu, H. (2016). Privacy Nudges for Mo-

bile Applications: Effects on the Creepiness Emotion

and Privacy Attitudes. In Proceedings of the 19th

ACM Conference on Computer-Supported Coopera-

tive Work & Social Computing, CSCW ’16, pages

1676–1690. Association for Computing Machinery.

ICISSP 2023 - 9th International Conference on Information Systems Security and Privacy

676