Should User-generated Content be a Matter of Privacy Awareness?
A Position Paper
Nicol
´
as Emilio D
´
ıaz Ferreyra, Rene Meis and Maritta Heisel
RTG User-Centred Social Media, University of Duisburg-Essen, Germany
Keywords:
Adaptive Privacy, Self-disclosure, Awareness, Social Network Sites, Data Visceralization.
Abstract:
Social Network Sites (SNSs) like Facebook or Twitter have radically redefined the mechanisms for social
interaction. One of the main aspects of these platforms are their information sharing features which allow
user-generated content to reach wide and diverse audiences within a few seconds. Whereas the spectrum of
shared content is large and varied, it can nevertheless include private and sensitive information. Such content
of sensitive nature can derive in unwanted incidents for the users (such as reputation damage, job loss, or
harassment) when reaching unintended audiences. In this paper, we analyse and discuss the privacy risks of
information disclosure in SNSs from a user-centred perspective. We argue that this is a problem of lack of
awareness which is grounded in an emotional detachment between the users and their digital data. In line with
this, we will discuss preventative technologies for raising awareness and approaches for building a stronger
connection between the users and their private information. Likewise, we encourage the inclusion of awareness
mechanisms for providing better insights on the privacy policies of SNSs.
1 INTRODUCTION
In 1966, tobacco companies across the United States
were affected by a law that later on changed the stan-
dards for the commercialization and distribution of
cigarettes. For the first time in the history, a legis-
lation requiring warnings about the risks associated
with the consumption of tobacco was proposed by the
U.S Congress (Hiilamo et al., 2014). Since then, the
companies began fighting against Health Warning La-
bels (HWLs) in cigarette packs basically arguing that
people already knew the hazards of smoking. Despite
their efforts on blocking or weaken HWLs, nowadays
many countries have included and implemented HWL
in their legislations (Hiilamo et al., 2014).
Social Network Sites (SNSs) are spaces which
are not free of privacy risks, and like in the case of
cigarettes consumers, users of SNSs might have heard
about some of these risks before or during their activ-
ity period (i.e. before or after opening an account in a
SNS). While one might argue that the risks of disclos-
ing personal or sensitive information in SNSs are not
as severe as the risks of smoking, unwanted incidents
such as job loss, reputation damage, or unjustified dis-
crimination should not be neglected or disregarded.
However, very little information (for not to say none)
is provided by the SNSs about the potential risks of
information sharing.
Privacy policies can be considered as an initial ap-
proach towards the information on potential privacy
risks. However, these electronic documents are shown
once to the users (when registering), and are hardly
revised by them in the future. Moreover, privacy poli-
cies basically inform about which data is collected,
how is processed, and under which conditions it is
disclosed to third parties; without any emphasis on
informing about potential risks. If we add to this that
users are not strongly attached to their private infor-
mation, then the chances of users regretting to have
shared private information increases.
We believe that, like tobacco consumers, the
users of SNSs should be empowered with informa-
tion about the potential risks of information sharing.
Moreover, we believe that awareness mechanisms can
be a good alternative not only to inform the users
about such risks, but also to create a stronger tie be-
tween them and their private information. In this
paper we take a closer look at the privacy risks as-
sociated with user-generated content in SNSs in or-
der to discuss possible solutions to this issue. More-
over, we provide arguments towards the use of adap-
tive preventative technologies to move towards a more
az Ferreyra N., Meis R. and Heisel M.
Should User-generated Content be a Matter of Privacy Awareness? - A Position Paper.
DOI: 10.5220/0006517302120216
In Proceedings of the 9th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (KMIS 2017), pages 212-216
ISBN: 978-989-758-273-8
Copyright
c
2017 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
privacy-aware social environment in SNSs.
The rest of the paper is organized as follows. In
the next section, we discuss the motivation scenario
and the paper’s background. In Section 3, we analyse
the role of privacy policies and media technologies on
modulating users’ perceptions towards their private
data. Following, we discuss in Section 4 preventative
technologies for the generation of awareness within
SNSs. In Section 5, we analyse an approach for in-
corporating privacy heuristics derived from users’ re-
grettable experiences into the design of preventative
technologies. Thereafter, we discuss the advantages
and drawbacks of this approach in Section 6. Finally,
we conclude in Section 7 with an outlook and consid-
erations for further research.
2 MOTIVATION AND
BACKGROUND
In 2018, the EU’s new General Data Protection Reg-
ulation (GDPR) (Regulation, 2016) will come into
force as the conclusion of a hectic debate which has
involved academics, Internet service providers and in-
ternational organizations across the world. For many,
the Internet is considered an open platform for demo-
cratic participation which promotes freedom of ex-
pression and the right to information access. There-
fore, is not surprising that the GDPR, and more
specifically the Right to be Forgotten (RTBF), raises
concerns related to abusive removal demands of user-
generated content (e.g. public officers trying to sup-
press criminal records), and other issues about poten-
tial unjustified censorship
1
. This is a debate which
mainly circles around the right to erase or de-list in-
formation put online by another Internet user. In this
work, we do not aim to discuss this aspect of the
GDPR. Instead, we look to resume the discussion to
the information that users disclose about themselves
in SNSs.
One of the critical concepts included in the GDPR
is the one of “personal data”. For instance, Article 4
says that information related to a “data subject” (i.e.
an identified or identifiable natural person
2
), such as
name, identification number, location, factors specific
to his/her physical, physiological, genetic, mental,
economic, cultural or social identity, should be con-
sidered as personal information and therefore require
1
The work by Keller (Keller, 2017), offers a clarifying
view on the RTBF and its hazards for freedom of expression
and information rights on the Internet.
2
An identifiable person is one who can be identified, di-
rectly or indirectly.
unambiguous consent to be processed. Likewise, Ar-
ticle 9 says that racial or ethnic origin, political opin-
ions, religious or philosophical beliefs, or trade-union
membership should not be processed unless the data
subject gives explicit consent. Unambiguous consent
can be given through a conduct that clearly indicates
that the data subject agrees with the proposed process-
ing of his/her personal data (e.g. when telling the doc-
tor about the medical ailment one is suffering while
he/she enters notes in a computer system). On the
other hand, explicit consent should be given trough
an explicit action by the data subject. This is normally
granted after the data subject clicks on ‘Yes, I agree”
in the privacy policies of the service provider.
3 DATA “VISCERALIZATION”
One of the main reasons for differentiating private
data from general data are the risks associated with
their inappropriate processing and public disclosure.
Basically, the GDPR encloses an implicit warning
message for data processors (the SNSs in our case)
which is that they should safeguard data subjects from
unwanted incidents (such as unjustified discrimina-
tion, political or religious persecution, or fraud) by
treating carefully their personal data. Privacy policies,
on the other hand, inform the users about which infor-
mation will be collected, processed, used, disclosed
and managed by the data processor. As one can ob-
serve, there is a semantic difference in the message of
the GDPR and the one of privacy policies. Whereas
the GDPR endows service providers with a better per-
ception on the importance of the users’ personal infor-
mation, privacy policies do not provide cues to data
subjects about the importance of their own personal
information. Consequently, privacy policies in some
point modulate users’ perceived severity of privacy
risks in SNSs.
Like privacy policies, information sharing inter-
faces of SNSs also play an important role in shaping
our perceptions of information privacy (Stark, 2016;
D
´
ıaz Ferreyra et al., 2017a). Such interfaces are the
entry point of user-generated content which, in many
cases, contains private information. However, since
digital data is intangible, information sharing inter-
faces of SNSs regulate users’ emotional perception
and attachment towards their private information. Let
us consider the following example: imagine that a
stranger stops you in the street and asks you for your
passport. It is quite unlikely that someone would
grant this request in the real world. Moreover, this
situation would normally come along with a visceral
reaction (i.e. an instinctive gut-deep bodily response
like a burning sensation in the stomach) as conse-
quence of this unexpected request. However, when
this information is requested through the interfaces of
a SNSs, such reactions do not seem to arise. Conse-
quently, privacy policies and sharing interfaces are not
succeeding in taking the users’ emotional perception
of their private data to the visceral level.
4 PRIVACY AWARENESS IN SNSs
Like in the case of HWL for the commercialization of
cigarettes, awareness mechanisms for SNSs can con-
tribute to bridge the emotional gap between users and
their digital data. In this section we discuss different
preventative technologies oriented to generate aware-
ness in online self-disclosure scenarios. This is, sce-
narios in which users intend to reveal their own pri-
vate information in SNSs.
4.1 Preventative Technologies
Different preventative technologies have been pro-
posed for mitigating the unwanted consequences of
online self-disclosure (Calikli et al., 2016; D
´
ıaz Fer-
reyra et al., 2016; Fang and LeFevre, 2010; Wang
et al., 2013; Ghazinour et al., 2013). One of the most
representative of these approaches is the one by Wang
et al. (Wang et al., 2013) consisting of three plugins
for Facebook. These plugins called “privacy nudges”
intervened when the user was about to post a message
in his/her Facebook biography either (i) introducing
a delay, (ii) providing visual cues about the audience
of the post, or (iii) giving feedback about the mean-
ing (positive or negative) of the post. However, since
the feedback generated by the nudges was the same
for every user of Facebook, they did not succeed on
reaching high levels of acceptance. This is, some
users liked them and others found them annoying.
Consequently, this type of technology should provide
adaptive feedback and awareness to their users in or-
der to being widely adopted.
4.2 Adaptive Awareness
Adaptive preventative technologies seek to develop
mechanisms capable to provide tailored feedback and
awareness to their users. One of the preventative
technologies which follow this direction is the one of
Ziegeldorf et al., consisting in a framework of per-
sonalized privacy metrics for the generation of adap-
tive awareness (Ziegeldorf et al., 2015). This ap-
proach, called Comparison-based Privacy (CbP), con-
sists of analysing different comparison metrics which
are computed over the content being shared among
different comparison groups. Basically, comparison
groups consist of groups of people with which the
user can intuitively relate to (e.g. family, friends and
colleagues, users with the same profession or same
age). Likewise, comparison metrics capture aspects
of the sharing behaviour within a privacy group, such
as the sentiment and the type of the content being
shared. A user can choose for instance to compare
the amount of hate speech in his/her posts against the
one of people with his/her same profession. If this
value exceeds a given threshold, then the system alerts
the user. Thresholds can be set individually by users,
or according to general profiles representing differ-
ent privacy attitudes (e.g. unconcerned, pragmatist or
fundamentalist). Approaches like this one overcome
the engagement issue caused by generic warning mes-
sages of static approaches.
5 VISCERAL-AWARENESS
DESIGN
One of the key elements of HWLs in cigarettes pack-
aging is that they include pictorial representations of
the risks of tobacco consumption. This is done in or-
der to make users perceive such risks in a more vis-
ceral way. In the case of online self-disclosure, regret-
table experiences come along with visceral reactions
from the users. This is, when a user lives an unwanted
incident after disclosing personal data in SNSs, then
a feeling of regret and repentance arises together with
a visceral reaction. In this section, we discuss design
principles introduced by D
´
ıaz Ferreyra et al. to in-
clude regrettable experiences into the design process
of preventative technologies.
5.1 Privacy Heuristics
D
´
ıaz Ferreyra et al. propose to take into ac-
count regrettable self-disclosure experiences in or-
der to endow preventative technologies with visceral-
awareness principles (D
´
ıaz Ferreyra et al., 2017a).
Basically, they suggest that privacy heuristics (best
practices) can be derived from regrettable self-
disclosure experiences and used thereafter to raise pri-
vacy awareness. For this, they introduce a Privacy
Heuristics Derivation Method (PHeDer) for eliciting
privacy best practices from user’s regrettable expe-
riences (D
´
ıaz Ferreyra et al., 2017a). The first step
of this method, called Regret Acknowledgement, con-
sists on gathering evidence about a regrettable expe-
rience and describe it in terms of: (i) the informa-
tion that was disclosed (ii) the unintended audience it
reached, and (iii) the unwanted incidents that lead the
user to a feeling of regret. The output of this step can
be represented as in Fig. 1, where the user reported
to have shared his/her political affiliation in a public
post. Once the regrettable scenario is described, it is
forwarded to the next step called Regret Analysis.
The Regret Analysis step consists of refining the
scenario of Fig. 1 into privacy risks consisting of a
7-tuple of elements: a list of personal attributes, the
unintended audience, the unwanted incident, the fre-
quency of the unwanted incident, the impact of the
unwanted incident, the risk level, and the user’s pri-
vacy attitude. Whereas the personal attributes can
be derived from articles 1 and 9 of the GDPR, fre-
quency and impact of the unwanted incident (and con-
sequently the risk level) can be expressed using nom-
inal scales. Privacy attitudes are one of the pragma-
tist (medium privacy concern), fundamentalist (high
privacy concern), or unconcerned (low privacy con-
cern). For the example of Fig. 1, the output of this step
would be risk([political opinion], work colleagues,
wakeup call, likely, major, very high, pragmatist).
This information if then forwarded to the third step
of the method which is Heuristic Design.
USER’S POST
“Seriously? Trump became president? What is
happening to the world!? #republicanssuck”
Actual Audience: PUBLIC.
Unintended Audience: The user’s work col-
leagues.
Unwanted Incidents: Wake-up call from superior.
Figure 1: Example of self-disclosure scenario.
The Heuristic Design step uses Constraint Based
Modeling (CBM)(Mitrovic and Ohlsson, 2006) as the
design principle for encoding the outcome of step 2
into a privacy heuristic. In CBM a constraint con-
sists of a pair of relevance and satisfaction condi-
tions, where each member of the pair can be seen as
a set of features or properties that a disclosure sce-
nario must satisfy. For the given example, the rele-
vance conditions would be the existence of a political
opinion inside a post, and the satisfaction condition
would be not include the work colleagues in the post
audience. Such constraints can be expressed using
Horn clauses in Prolog and be included in the final
step (Constraint Integration) in the Privacy Heuris-
tics Data Base (PHDB) of an Instructional Awareness
System (IAS) (D
´
ıaz Ferreyra et al., 2017a; D
´
ıaz Fer-
reyra et al., 2016).
5.2 Instructional Awareness
Basically, an IAS uses the heuristics inside a PHDB
to detect potentially regrettable disclosures. Such
heuristics are evaluated when a “post” event takes
place in order to determine if the disclosure action can
derive in a regrettable scenario for the user. This is,
done first by evaluating the relevance condition of the
heuristics. Let us consider a scenario where a user
wants to disclose once again his/her political affilia-
tion inside a public post. Let us also consider that the
heuristic discussed in Section 5.1 is part of the PHDB.
In this case, IAS will detect that the disclosure can
lead to a potential regret, and therefore proceeds to
raise a warning to the user.
Figure 2: Instructional Awareness System (IAS).
In order to generate adaptive feedback, IAS takes
into account adaptivity variables such as the user’s
privacy attitude, the number of times the user has ig-
nored/accepted the warnings, and how often he/she
discloses private information. This information is
stored in a User Performance Data Base (UPDB) that,
together with the PHDB, makes up IAS’s Knowledge
Base (KB). With the information stored in its KB, IAS
can generate a message such as “Revealing your polit-
ical affiliation to your work colleagues can bring you
problems. Do you want some hints on how to pro-
tect your private data?” and recommend the user to
restrict the post’s audience (for instance to “friends
only”). Since information about the risks is also kept
in the KB, an IAS can also provide such additional
information in the warning message.
6 DISCUSSION
Following a similar approach to the HWLs in
cigarettes packages, D
´
ıaz Ferreyra et al. propose to
inform the users of SNSs about the risks of online
self-disclosure though an IAS. For this, IAS requires
risk knowledge which is stored in a PHDB and ob-
tained through a privacy heuristics derivation method.
This method is an offline approach for eliciting pri-
vacy heuristics from regrettable online self-disclosure
experiences. Basically, the input of the method are
the experiences that users have reported themselves
(to the development team of IAS for instance), or the
outcome of an empirical research (e.g. questionnaires
or face to face interviews). This approach is effective
for building a baseline of heuristics prior to the exe-
cution of the system. However, eliciting new entries
of the PHDB requires the execution of this process
which can be expensive and inefficient in terms of the
resources and time needed to conduct interviews and
process the outcome of them. One way to overcome
this issue is to consider deleted posts with private in-
formation as potential sources of heuristics (D
´
ıaz Fer-
reyra et al., 2017b). This is, to use such posts as the
input of a machine learning engine for the automatic
derivation of privacy heuristics at runtime. This way,
the PHDB can be updated with new heuristics with-
out having to execute offline iterations of the PHeDer
method.
7 OUTLOOK AND CONCLUSION
Adaptive awareness technologies seem to be promis-
ing approaches for empowering the users of SNSs in
making wiser and more informed decisions, as to pro-
tect them from the risks of over-sharing private infor-
mation. We believe that this is not a minor issue that
should be taken seriously into consideration by Inter-
net service providers, multilateral organizations and
policy makers. We have used the example of HWLs in
cigarettes packages as a motivating scenario for work-
ing towards a more privacy aware social environment
in SNSs. Certainly, this topic will be part of a in-depth
and intense debate in the future. Therefore we hope
this paper will offer a more clarifying view on this is-
sue and serve as an instrument for the development of
more effective solutions.
ACKNOWLEDGEMENTS
This work was supported by the Deutsche
Forschungsgemeinschaft (DFG) under grant No.
GRK 2167, Research Training Group ”User-Centred
Social Media”.
REFERENCES
Calikli, G., Law, M., Bandara, A. K., Russo, A., Dick-
ens, L., Price, B. A., Stuart, A., Levine, M., and Nu-
seibeh, B. (2016). Privacy Dynamics: Learning Pri-
vacy Norms for Social Software. In Proceedings of the
11th International Symposium on Software Engineer-
ing for Adaptive and Self-Managing Systems, pages
47–56. ACM.
D
´
ıaz Ferreyra, N. E., Meis, R., and Heisel, M. (2017a). On-
line Self-disclosure: From Users’ Regrets to Instruc-
tional Awareness. In Proceedings of the IFIP Interna-
tional Cross-Domain Conference (CD-MAKE). Ac-
capted for publication.
D
´
ıaz Ferreyra, N. E., Meis, R., and Heisel, M. (2017b). To-
wards an ILP Approach for Learning Privacy Heuris-
tics From Users’ Regrets. In Proceedings of the 4th
European Network Intelligence Conference (ENIC).
Accapted for publication.
D
´
ıaz Ferreyra, N. E., Sch
¨
awel, J., Heisel, M., and Meske, C.
(2016). Addressing Self-disclosure in Social Media:
An Instructional Awareness Approach. In Proceed-
ings of the 2nd ACS/IEEE International Workshop on
Online Social Networks Technologies (OSNT). AC-
S/IEEE.
Fang, L. and LeFevre, K. (2010). Privacy wizards for social
networking sites. In Proceedings of the 19th Inter-
national Conference on World Wide Web, WWW ’10,
pages 351–360, New York, NY, USA. ACM.
Ghazinour, K., Matwin, S., and Sokolova, M. (2013).
YourPrivacyProtector: A Recommender System for
Privacy Settings in Social Networks. International
Journal of Security, Privacy and Trust Management
(IJSPTM), 2(4).
Hiilamo, H., Crosbie, E., and Glantz, S. A. (2014). The
evolution of health warning labels on cigarette packs:
the role of precedents, and tobacco industry strategies
to block diffusion. Tobacco control, 23(1):e2–e2.
Keller, D. (2017). The right tools: Europe’s intermediary li-
ability laws and the 2016 general data protection regu-
lation. Technical report, Stanford Law School Center
for Internet and Society.
Mitrovic, A. and Ohlsson, S. (2006). Constraint-based
knowledge representation for individualized instruc-
tion. Computer Science and Information Systems
(ComSIS) Journal, 13(1).
Regulation, G. D. P. (2016). Regulation (EU) 2016/679
of the European Parliament and of the Council of 27
April 2016 on the protection of natural persons with
regard to the processing of personal data and on the
free movement of such data, and repealing Directive
95/46/EC (General Data Protection Regulation). Offi-
cial Journal of the European Union (OJ), 59:1–88.
Stark, L. (2016). The Emotional Context of Information
Privacy. The Information Society, 32(1):14–27.
Wang, Y., Leon, P. G., Scott, K., Chen, X., Acquisti, A., and
Cranor, L. F. (2013). Privacy Nudges for Social Me-
dia: An Exploratory Facebook Study. In Proceedings
of the 22nd International Conference on World Wide
Web, pages 763–770. ACM.
Ziegeldorf, J. H., Henze, M., Hummen, R., and Wehrle,
K. (2015). Comparison-based privacy: nudging pri-
vacy in social media (position paper). In International
Workshop on Data Privacy Management, pages 226–
234. Springer.