Ethics and Information Technology

https://doi.org/10.1007/s10676-018-9496-y

ORIGINAL PAPER

Serendipity as an emerging design principle of the infosphere:

challenges and opportunities
Urbano Reviglio1

© Springer Nature B.V. 2019

Abstract
Underestimated for a long time, serendipity is an increasingly recognized design principle of the infosphere. Being influ-
enced by environmental and human factors, the experience of serendipity encompasses fundamental phases of production,
distribution and consumption of information. On the one hand, design information architectures for serendipity increases
the diversity of information encountered as well as users’ control over information processes. On the other hand, serendipity
is a capability. It helps individuals to internalize and adopt strategies that increase the chances of experiencing it. As such,
the pursuit for serendipity can help to burst filter bubbles and weaken echo chambers in social media. The article reviews
the literature on emerging issues surrounding serendipity in human–computer interactions. By doing so, it firstly presents
the study of serendipity and the debate about its role in digital environments. Then, it introduces the main features of a pre-
liminary architecture for serendipity. Finally, it analyzes from an interdisciplinary perspective the values that embraces and
sustains. The conclusion is that serendipity can be conceived as an emerging design and ethical principle able to strengthen
media pluralism and other emerging human rights in the context of online personalization. Main limitations and potential
unintended consequences are also discussed.

Keywords Serendipity · Design ethics · Nudging · Personalization · Filter bubbles · Echo chambers

Introduction information as to feed users with highly personalized infor-

Information is power, and it is increasingly essential for both
individuals and societies. In particular, the capacity to man-
age knowledge is a fundamental skill to thrive in the digital
era. Yet, nowadays such potential is not fully exploited . On
the one hand, individuals are overloaded by information.
While media landscape rapidly changes, the majority of
individuals rarely have sufficient (digital) literacy to respon-
sibly deal with such epistemological disruption. Greater
knowledge, in fact, doesn’t always bring with it greater
understanding. Radicalization and political polarization can
arise from consuming information receptively and acritically
(Lynch 2016). On the other hand, trans-national oligopolis-
tic information intermediaries—in particular Facebook and
Google—shape the flows of information, so does social life.
Their business model threatens the diversity and richness of
* Urbano Reviglio
urbanoreviglio@hotmail.com
1
Law, Science and Technology (ERASMUS+), University
of Bologna, Bologna, BO, Italy
mation mainly based on popular and homophilous content,
more emotional than intellectual, and efficiently designed to
engage rather than to learn and/or educate.
Online personalization performs a fundamental role of
knowledge management in order to restrain information over-
load, reduce complexity and, ultimately, satisfy individuals.
At the same time, it also sustains problem-solving and deci-
sion-making processes to preserves our scarcest resources:
time and attention. Yet, personalization of media content
can create filter bubbles and strengthen echo chambers that
restrain the exposure to diverse, challenging and serendipitous
information (Pariser 2011; Sunstein 2017). This represents a
fundamental issue for media law and ethics both seeking to
preserve autonomy of choice and media pluralism in demo-
cratic societies. As a result, individuals might reduce their
informational empowerment and societies might become
more politically polarized.
Even though research about the detrimental effects of
personalization is often inconsistent, there is no doubt that
in the long run the algorithmic capacity to steer our lives in
increasingly sophisticated ways will dramatically expand.

13
Vol.:(0123456789)

Key questions need to be discussed further; for instance, to
what extent can information filtering account for the com-
plexity of individual identity? What are the main values and
trade-offs that inform designers in such a fundamental soci-
etal algorithmic arbitrage? And, eventually, what might be
the role of serendipity in this context?
The paper provides a theoretical investigation of the phe-
nomenon of serendipity to answer such questions. For this
purpose, it draws on insights from media and information
ethics and Science and Technology Studies (STS) to argue
that, although often used in the mainstream as a vague buz-
zword, serendipity is a fundamental design principle for
socio-technical systems and, at the same time, it represents an
ethical horizon for both individuals and democratic societies.
Thus, an analysis of the dynamics of designing for serendip-
ity and an ethical overview of the values that it underpins will
be provided. No extensive literature review and ethical analy-
sis of the relationship between serendipity and filter bubbles
in social media has indeed been done so far.
The paper is organized as follows. Firstly, we present the
concept of serendipity and its study, its preconditions and
its outcomes. Secondly, we briefly discuss the role of seren-
dipity in digital environments, how personalization threat-
ens it, and how to cultivate an information architecture for
serendipity. Thirdly, we explore serendipity and the values
it embraces and sustains, both from a philosophical and an
educational perspective. Finally, we discuss the limitations
and the unintended consequences of conceiving serendipity
as a design principle.
Serendipity and its study
Serendipity is the art of discovering new things by observ-
ing, and learning from encountering unexpected informa-
tion. It has received attention in several fields from sociology
of science to epistemology, from psychology and innovation
studies to information and computer science. It is no surprise
that there is no wide consensus on its meaning. It is indeed
widespread to consider serendipity as a romantic ill-defined
buzzword (Makri 2014). Serendipity is relatively a neolo-
gism with no equivalent in languages other than English.1 It
originated from Horace Walpole in 1754, that was inspired
by the Persian fairy tale “The Three Princes of Serendip”,
which narrates how these traveling princes were “always
making discoveries, by accidents and sagacity, of things
which they were not in quest of”. Largely ignored since then,
it gained popularity from mid-1900s when it was applied
to various breakthroughs in scientific research, until voted
1
It had been indeed perceived as an “esoteric word” given that it did
not appear in any of the abridged dictionaries until 1951 (Merton and
Barber 2006).
13
U. Reviglio
as “UK’s favorite word” in 2000 (Rubin et al. 2011). It is
today a rather known concept often defined as “an unex-
pected experience prompted by an individual’s valuable
interaction with ideas, information, objects, or phenomena”
(McCay-Peet and Toms 2013).
As serendipity became famous, “the vogue word became
a vague word” (Merton and Barber 2006). Serendipity is
indeed often incorrectly referred to as the “happy accident”,
thus interpreted as mere coincidence, luck and providence.
This dilution of meaning challenges its application in
research (McCay-Peet and Toms 2017). In fact, serendipi-
tous discoveries can be perceived as accidentals but they are
usually the result of groundwork, observation, and previ-
ous knowledge. As Louis Pasteur famously claimed: “in
the field of observation, chance favours only the prepared
mind.” (Merton and Barber 2006).2 Thus, serendipity has to
be considered also as a capability (de Rond 2014), intended
as being capable of making use of the opportunities that
the environment provides (Sen 2005). As various definitions
assign different weights to personal and environmental fac-
tors, between individual sagacity and lucky chance, this ten-
sion, eventually results in a combination of them, differing
case by case.
In the context of digital environments, serendipity can be
studied under different perspectives. For information behav-
ior research it has been approached as a quality of someone
or, in research relating to recommender systems and search
engines, as a quality of an event or something. While being
approached as either a personal proclivity or a phenomenon
or event, the most holistic approach to serendipity occurs in
information science and human–computer interaction (HCI)
research where serendipity is often approached as an expe-
rience or a process. As such, it has various components,
dimensions and environmental factors (Kotkov et al. 2016)
as well as affordances (Bjorneborn 2017).
It is possible to generalize serendipitous encounters into
three-main types (McCay-Peet and Toms 2017)3:
2
Consider one of the most famous example of serendipity: the dis-
covery of penicillin by Fleming. It has been argued that at least 28
scientists before Fleming reported the same colonies of bacteria that
led him to the discovery of penicillin (de Melo 2018). Yet, all chose
to view that anomaly as an unfortunate error rather than an opportu-
nity for discovery.
3
There are many other taxonomies of course (for a recent one see
Yaqub 2018). Yet, one of the most popular and intuitive taxonomy of
serendipitous discoveries is the one by Friedel (see de Melo 2018),
which identifies three main forms from science’s historical exam-
ples: Columbian, which occurs when one is looking for one thing of
value, but finds another one (from Columbus’s unsought discovery of
America); Archimedean, which occurs when one discovers sought-for
results, although by routes not logically deduced but luckily observed
(from Archimedes’s “Eureka” moment); and Galilean, which occurs
when one discovers something valuable without intentionally seeking
it (from Galileo’s unexpected discoveries through his telescope).
Serendipity as an emerging design principle of the infosphere: challenges and opportunities
• Type A: from observations to a solution. It befalls when
people are not looking for any information in particular.
For instance, while scrolling a social media’s feed. This
also represents an example of the opportunistic acquisi-
tion of information (OAI), so common in digital environ-
ments (Erdelez 1997).
• Type B: from problem I to a solution to problem II. It
occurs when people are looking for information on a cer-
tain problem and find a different but previously unsolved
problem.
• Type C: unexpected solutions, also called pseudo-seren-
dipity. It usually befalls when an individual is seeking
information purposefully but accidentally finds an unre-
lated, unexpected one.
As it is a subjective experience—any individual has
different strategies to experience different kinds of seren-
dipity—in this paper we broadly intend serendipity as the
experience depending on both environmental conditions
and individual capability to accidentally encounter—both
opportunistically (Type A) and purposefully (Type B and
C)—meaningful and unexpected information that may help
to solve a problem, make you discover a novel idea, interest,
or even change a belief.4
The ability to extract knowledge serendipitously covers
all areas of human activity, including ‘science, technology,
art, and daily life’ (van Andel 1994). According to Merton
and Barber (2006) who wrote the most extensive research
about its meaning and diffusion, serendipity is one of the
main forces that has steered the progress of science. It is
full of renowned examples: from penicillin to DNA, from
string theory to mirror neurons.5 The role of serendipity on
epistemology of science is thus well-established. There is,
however, a lack of advancement in this field. Interestingly,
Campanario (1996) suggests that the reason for this lies in the
mythization of scientific research and the difference between
the ‘context of discovery’ and the ‘context of justification’.
Science is indeed supposed to be something that is totally
under control, so that scientists may be reluctant to admit that
the discoveries for which they are honoured were accidental.
Similarly, in everyday life the role of chance and ser-
endipity is often underestimated. Though relatively less
momentous, serendipitous insights occur on a daily basis,
particularly on the Internet. Accidental encountering is
4
Actually, there is no consensus on the definition of serendipity,
and it also varies on the field of study (McCay-Peet and Toms 2017).
Moreover, some of the adjectives used are difficult to operationalize it
are often used interchangeably (for instance, unexpectedness and sur-
prise, or usefulness, interestingness and meaningfulness).
5
Nearly 1 in 10 of the most-cited scientific papers of all time explic-
itly mention serendipity as a contributing factor (Campanario 1996)
and it has been estimated that over 50 percent of scientific discovery
might have been unintended (Dunbar and Fugelsang 2005).
indeed an integral part of everyday information behavior
(Erdelez 2004), also called micro-serendipity (Bogers and
Björneborn 2013). Even though serendipity studies have
suggested that serendipity is a rare, elusive phenomenon, in
today’s information societies for many serendipity seems so
common that they do not always reflect upon it and therefore
not even recognize it.
Most of the studies on serendipity focus particularly on
scientific discovery rather than on everyday serendipity.
Extensive accounts on how to research serendipity (McCay-
Peet and Toms 2017) and cultivate it in digital environments
(Race and Makri 2016; Bjorneborn 2017; de Melo 2018)
provided the ground for novel studies. These literature
reviews suggest how the tension between the need to man-
age both the quantity and quality of information has been a
key driver of these research. They show how serendipity in
digital environments can be reduced and cultivated through
design, affordances and recommender systems.
Furthermore, it is researched the increasing role of ser-
endipitous news consumption online (or Incidental Expo-
sure to Online News, IEON) for the formation of individual
and public opinion (Yadamsuren and Erdelez 2016). Indeed
media researchers have largely ignored such phenomenon
for a long time, probably due to the difficulty in capturing
it. In recent years, however, more studies started to explore
how individuals obtain information accidentally and oppor-
tunistically, rather than mostly purposively.
Discussing serendipity in digital environments
Fathers of cybernetics6 recognized the value of a necessary
(but not sufficient) element of serendipity: surprise, which
seems to operate as a cognitive/emotional reaction to ser-
endipitous encounters (Rubin et al. 2011). In mathemati-
cal terms, Shannon (1948) argued that information is pro-
portional to its deviation from expectation. In other words,
the amount of information is a measure of surprise.7 From
a theoretical perspective, instead, the question of whether
computers are able to ‘take us by surprise’ was famously
raised by Alan Turing (1950). Given the increasing progress
6
Interestingly, it is possible to trace back the influence of serendip-
ity in the foundations of cybernetics by Norbert Wiener. Olma (2016)
argues that the Rad Lab—the famous Radiation Laboratory located at
the Massachusetts Institute of Technology (MIT)—was an example of
an ‘institutionalised serendipity’ environment which created the nec-
essary conditions that encouraged “a transversal exchange of knowl-
edge that, in turn, enabled Wiener to create the discipline of cybernet-
ics, which itself allowed for the development of ARPANET, one of
the technical foundations of the internet” (de Melo 2018).
7
If we want to generalize the concept of surprise we may have that
an event which occurs with high probability should have a low sur-
prise, whereas an event which occurs with low probability should
have a high surprise. This sheds light on the paradoxical challenge to
program serendipity.
13

of and interaction with machine learning systems, this latest
issue has particularly debated nowadays.
There are generally two main intertwined literature in
computational creativity on the subject (Colton and Wig-
gins 2012; Delacroix 2018); on the one hand, one focuses
on improving the learning performance of the system
through the ability to surprise itself while, on the other hand,
research is concerned with introducing serendipitous outputs
for the end-users. Scholars indeed argue that it is possible
to proactively design algorithmic systems that may be sur-
prised and that provide non-trivial surprises, in other words
‘artificial serendipity’ (Delacroix 2018).
Serendipity can be explicit and implicit (de Melo 2018).
Not only it can be intentionally designed but it is more often
a by-product of digital environments. Media theorist Steven
Johnsons described the Internet as ‘the greatest serendip-
ity engine in the history of culture’ (Carr 2016). Every day
Internet users can in fact discover plenty of surprising and
meaningful information that can change their current task,
their own beliefs, or even their lives. Hyperlinked digital
environments are indeed a fertile ground for serendip-
ity (Thurman and Schifferes 2012). Social media, search
engines, RRSs, social bookmarking such as folksonomy
(Auray 2007),8 and more generally browsing, are power-
ful sources of coming across information serendipitously.9
Mainstream services like Netflix, Amazon, Google Now and
Spotify also provide serendipitous recommendations. Yet
they tend to exploit more convergent systems—as the capac-
ity to discover the right thing at the right time, to cater to
the user’s perceived intentions, interests, tastes—rather than
divergent systems—as the increasing of information diver-
sity in order to expand user’s horizons and to help uncover
surprising discovery (de Melo 2018).
Indeed, Internet can be ever more serendipitous, especially
in social networking sites (SNS) and video sharing platforms
such as Youtube when political content is encountered and,
eventually, consumed. While in newspapers, and more gener-
ally in traditional media, any procedure to select and prior-
itize information to some extent recognizes and seeks to solve
in a beneficial manner the tension between a generalized rel-
evance—what individuals are expected to want—and seren-
dipity—what individuals may like—due to personalization
8
Folksonomy, also known as collaborative tagging, social classifica-
tion, social indexing, and social tagging, is a collaborative user-gen-
erated system of classifying and organizing online content into differ-
ent categories by the use of metadata such as electronic tags. Famous
social networks based on it are del.icio.us and StumbleUpon.
9
Even the metaphor ‘surfing the Internet’ was chosen to refer to a
fun feeling and “something that would evoke a sense of randomness,
chaos, and even danger” (Polly 1992). Without search engines, the
journey on the web was indeed initially intended as discovering what
was out there—accidentally—not on finding specific content (Hend-
ler and Hugill 2013).
13
U. Reviglio
in digital environments this balance shifted from serendipity
to relevance (Thurman 2011). Under the current business
model, however, relevance and quantity tends to outstrip
serendipity and quality, providing only an ‘illusion of seren-
dipity’ (Erdelez and Jahnke 2018).10 Mainstream platforms
do not actively provide information that actually lies beyond
people’s biases and tastes. In social media, most users pas-
sively consume personalized content designed for hedonism,
entertainment, sensationalism, self-reassuring or subtle self-
propaganda.11 So far, the capacity to control curation and
to navigate filters is rather limited (Harambam et al. 2018).
Aside from natural human proclivities such as selective
exposure, confirmation bias and homophily, personalization
of media content can limit information discovery, particularly
if implicit.12 As a consequence, filter bubbles (Pariser 2011)
and echo chambers (Sunstein 2009) are strengthened. These
represent two sides of the same token. While the first refers to
the fact that a user continue to receive and consume content
that reinforces its opinions and interests, the latter refers to a
group situation where information is uncritically spread and
amplified, whereas dissonant information is often ignored.
These reduce serendipitous discoveries, and such loss would be
indeed more salient in SNS where users are in a more hedonic
rather than purpose-driven mind-set (Lutz et al. 2017).13
10
See for example the Algotransparency.org project which shows
how certain video on Youtube—mostly about conspiracy theories—
are much more recommended than others. One of its founder—a for-
mer Google employee—argues the possibility that certain videos that
discredit traditional media are more recommended with the goal of
further engage users within its platform. See https://medium.com/@
guill aumechaslot/how-algor ithms -can-learn -to-discr edit-the-media
-d1360157c4fa.
11
Algorithms which predict individual’s preferences tend to nudge
users’ comfort zones. For instance, Facebook is deeply committed
to maintain friends’ relationships. Its newsfeed is therefore moder-
ated by homophily (DeVito 2017) which is, however, the primary
driver of content diffusion, especially misinformation and conspiracy
theories, with a frequent result of homogeneous, polarized clusters
(Del Vicario et al. 2016). Moreover, information intermediaries may
increase engagement also by developing unconscious addictive ritu-
als based on gamification and dark patterns with the help of affective
computing and “captology” (Fogg et  al. 2002). Algorithms indeed
explore manipulative strategies that may be detrimental to users
(Albanie et al 2017).
12
Implicit personalization determines user preferences from data
collected (Thurman and Schifferes 2012). It actually increases politi-
cal selective exposure as it makes information avoidance less psycho-
logically costly (Dylko et al. 2018).
13
To give a general portray of the magnitude of the phenomenon,
consider that the average time currently spent on Facebook by a user
(in total circa 2 billion) is about 1 h a day, and the posts encountered
are circa 350, prioritized on about 1.500 (thus, roughly 75% are hid-
den) (Backstrom 2013). In U.S., two-thirds (67%) report that they get
at least some of their news on social media (Shearer and Gottfried
2017). Yet, just 14% of Facebook users believe ordinary users have
a lot of control over the newsfeed and only about 36% intentionally
tried to influence that (Smith 2018).
Serendipity as an emerging design principle of the infosphere: challenges and opportunities
In past discussions about Internet developments hopes
and fears around the concept of serendipity arose. It is
indeed already for a long time that such concern has led
some to question about “The End of Serendipity” (Gup
1997). Imagining our future digital lives Negroponte (1996)
expected that depending on time available, time of day, and
our mood we would have wanted lesser or greater degrees
of personalization ‘to experience the news with much more
serendipity’.14 Today mainstream social media satisfy users
with tools to further personalize but very few affordances for
divergent discoveries.15 Users also don’t have much influ-
ence on recommendation processes (Loepp et al. 2014).
They heavily rely on implicit personalization and this may
become a default choice in future personalization (Thurman
and Schifferes 2012).16
Profiling technologies that enable personalization actually
cannot produce or detect a sense of self but they can, how-
ever, influence a person’s sense of self (Hildebrandt 2009).
This may have serious consequences that led to some moral
panic on the erosion of fundamental human rights. In this
context, Meckel (2011) made an appeal to ‘save serendip-
ity’ as a fundamental experience for individual progress.
Identities might indeed be undermined by profiling algo-
rithms that are stuck in the past and that may force us into a
never-ending time-warp, dwelling in the status quo of past
preferences and desires. Eventually, users begin to want what
is suggested to them without even recognizing it, in what
Zarsky (2002) defines as the “autonomy-trap”. As a result,
the ‘aspirational self’ might be impeded to blossom. One
can disagree on such potential determinism of algorithms
but approximated knowledge that may impact individuals’
preferences without our conscious awareness is certainly
constructed. To what extent, then, the algorithmic repre-
sentation of us is—or may ever be—actually reliable?
Hildebrandt (2017) argues that the ‘multiplicity of the
self’ is not computable and that such ‘incomputable nature
of the self’ should be somehow protected. In fact, not only
no personal profile can ever entirely identify an individual,
but there are always many—and sometimes radically differ-
ent—ways of computing the same person. Philosophers are
14
Negroponte referred to two main concepts: “daily me” and “daily
us”. With the first he referred to personalized online news summa-
ries (tending to a convergent system), while with the latter to non-
personalized online news summaries (tending to a divergent system).
Of course, these are not two black and white different states but one
tends to move between them.
15
Notably, the most significant affordance that Facebook currently
provides is to select the option “Most recent” stories in the Newsfeed.
However, anytime you re-launch the site the setting spontaneously
reset itself by default to “Top Stories”—the classic curated newsfeed.
See https://www.facebook.com/help/community/question/?id=10152
476808951132 and also https://www.facebook.com/help/3716758463
32829.
indeed questioning the significance of preserving instances
of quasi-casualties in the era of big data (Stiegler 2017). In
the same vein, it is stressed the importance to design deci-
sion-support systems that sustain the unavoidable need for
moral change (Delacroix 2018). Designing for serendipity,
therefore, acknowledges the intrinsic hedonistic redundancy
and identity reductionism of online personalization and,
thus, it represents the attempts to overcome them.
As personalization is improving its accuracy and the
quantity of valuable information increases, there is an
increasing recognition of the value of serendipity. MIT
Media Lab director Zuckerman (2013) is a prominent expo-
nent of serendipity, particularly concerning cyber-balkani-
zation, and how the development of a cosmopolitan culture
may be limited due to georeferenced algorithms. In the next
years, he argues, tools that enable serendipity will become
as important as a utility as search engines and social media
are today. In effect, there are more and more attempts to
pursue serendipity, sometimes merely as a marketing tool,
sometimes as a genuine design goal, sometimes also as
information behaviour strategies to enhance online discov-
ery, in everyday searches as in academic scholarly research
(Maloney and Conrad 2016).
Skepticism about programming serendipity is also wide-
spread (Krotoski 2011; Carr 2016). Once you create an
engine to produce serendipity, one may claim, you destroy
its essence. If serendipity could be controlled, then an event
is no longer serendipitous, but predictable or reproducible.
Others, however, have argued against this apparent con-
tradiction noting that even though serendipity cannot be
fully controlled being characterized by chance, cultivation
and perception of serendipitous encounters can indeed be
enhanced (Race and Makri 2016). A related argument is that
serendipity is so widespread in digital environments that if
you expect it you would lose its surprising power. On the
contrary, research on serendipity in information behaviour
emphasize the expectation of positive or fortuitous outcomes
(Makri and Blandford 2012). As such, ‘expect the unex-
pected’ can actually result in more serendipitous insights
(van Andel 1994).17
Despite criticism, there is an increasing recognition
that design for serendipity has the potential to prevent the
threats of filter bubbles and echo chambers (Yadamsuren
and Erdelez 2016; McCay-Peet and Toms 2017). In gen-
eral, this implies to design digital environments that increase
16
Consider that Youtube’s recommendations already drive more than
70% of the time spent in the video sharing platform. See https://www.
cnet.com/news/youtube-ces-2018-neal-mohan/.
17
As the Greek philosopher Heraclitus (544–484  B.C.) famously
argued: “if you do not expect it, you will not find the unexpected, for
it is hard to find and difficult.”
13

the likelihood of encountering information one may (not)
like18—in opposition to individuals’ hedonism and plat-
forms’ pursuit of profit—so as to mainly avoid that past
behaviours pre-determine and excessively strengthen future
ones.
Serendipity is not in danger, democracy
and autonomy are
Personalization of media content raises serious ethical and
legal concerns. From an individual perspective, it might
reduce opportunities to self-determine, by reducing the
exposure to alternative points of view and serendipity (Par-
iser 2011). The consequences may be various: from the limi-
tation of personal creativity to a reduction in the ability to
build productive social capital (in particular weak ties, see
Granovetter 1977). More generally, the main consequence
to provide a ‘too familiar world’ is that our online life would
eventually shift from an intersubjective to a subjective one
(Keymolen 2016).
From a collective perspective, media pluralism may be
weakened by personalization, making people more politi-
cally polarized and vulnerable to censorship and propaganda
or, better, to self-censorship and self-propaganda (Sunstein
2017a). This, in turn, would help to spread misinformation
(Vicario et al. 2016) and erode interpersonal trust (Key-
molen 2016). Critics, however, argue that these are moral
panics, and that personalization might instead foster the
cultivation of “expert citizens” with stronger group identi-
ties (Harambam et al. 2018). Another prominent risk, how-
ever, remains growing ‘epistemic inequality’ (Lynch 2016).
Certain privileged group of users, that have higher (digital)
literacy, would be able to reach a good balance between ‘rel-
evance and serendipity’, and the recommender systems may
also adapt to their ‘serendipitous proclivity’. Instead, a larger
group of users would risk to be exposed only to a minimum,
qualitatively inferior, range of information.
In practice, the risks of filter bubbles are very hard to
prove and, eventually, to counteract (Zuiderveen Borgesius
et al. 2016). Most research, in fact, is often inconsistent and
inconclusive because they are generally survey-based or
based on a small or unsatisfactory sample. In the light of
the rapidly changing media landscape, then, many studies
become rapidly out-dated. There is indeed a crisis on the
study of algorithms (Bodo et al. 2017). Its functioning is
opaque and ‘black-boxed’ (Pinch and Bijker 1984; Pasquale
18
While serendipity is usually acknowledged as a pleasant experi-
ence, in this paper it is also valued the role of unpleasant encounters,
usually called zemblanity. This means that designing for serendipity
unavoidably implies also unpleasant encounters—albeit potentially
serendipitous.
13
U. Reviglio
2015) and their interpretability is not even clear (Albanie
et al. 2017). Also, users consider filtering mechanism as
neutral and actually few recognize them or attempt to opt-
out (Gillespie 2014).19 Thus, key issues remain unsolved:
to what extent personalization is detrimental, to what extent
are information intermediaries complicit and responsible,
whether they should be the target of a policy focus and,
eventually, what kind of intervention might be pursued.
As a reaction to the new media environment, media ethi-
cists and policy-makers began to discuss how to maintain an
exposure to diverse information. From a theoretical perspec-
tive, in fact, all the models of democracy consider the conse-
quences of filter bubbles problematic for particular different
reasons (Bozdag and Van den Hoven 2015). Personalization
can thus be analyzed from multiple perspectives, in particu-
lar media pluralism, human rights and data protection.
Media pluralism is achieved when users autonomously
enjoy a diverse media diet. The exposure to various sources
and content, also discussed as “diversity by design” (Hel-
berger 2011), seems insufficient to ensure actual experience
of media diversity. Even if media diversity online is gener-
ally more than in traditional media, such exposure does not
always end up in an actual experience of diversity (Hoffmann
et al. 2015). Cognitive and affective factors that drive Inter-
net users must also be considered. This requires to employ a
user-centric perspective and extend beyond the assumption
that supply diversity equals experience of diversity, and that
diversity of sources equals diversity of content. The value
of pluralism and diversity are necessary yet not sufficient
given users’ limited attention. In fact, how much exposure
to how many different contents and sources may be ever
considered ‘sufficient’? How the experience of diversity
as an ethical and, eventually, policy goal can eventually be
reconceptualised?
Pluralism as a normative principle remains vague and
under-theorized, and it is not a reliable indicator of a soci-
ety’s level of freedom, since it may create only the illusion
of content diversity (Karppinen 2008). In the digital age it is
indeed becoming less clear in which sense it is meaningful
to speak of media pluralism if the consumption is poten-
tially characterized by limitless choice. Information diversity
remains indeed a fundamental value for democratic societies
to sustain media pluralism but it does not fully convey the
disruption brought by technology (Helberger et al. 2016).
Thus, how serendipity can help to enrich the conception of
media diversity online?
On the one hand, it has been already argued that a core
function of traditional media is to provide ‘reliable surprises’
19
Only 24% were aware that Facebook prioritizes certain posts
and hides others from users’ feeds while 37% believed every post is
included in the newsfeed (Powers 2017).
Serendipity as an emerging design principle of the infosphere: challenges and opportunities
in order to balance familiarity and chaos (Schönbach 2007).
Eventually, a policy framework of “serendipity by design”
might help to tackle the challenges of experiencing diver-
sity (Reviglio 2017). In fact, it would imply diversifica-
tion of information and its interactive control (Bjorneborn
2017). Also, Sunstein (2017a) advocated an “architecture
of serendipity” as it would sustain ‘chance encounters and
shared experiences’ considered as preconditions for a well-
functioning democracy. Serendipity, in fact, is primarily
cultivated through strengthening weak ties (Ahmadi and
Wohn 2018). Thus, it can counteract homophily and, at the
same time, if intended as users’ empowerment in controlling
information, it promotes autonomy (Sunstein 2017c). More
generally, it could strengthen media pluralism, potentially
resulting in a stronger common base of facts and arguments.
On the other hand, serendipity, can be conceived as an
adaptive ethical mediator not only able to inform designers,
but also policy-makers, to eventually strengthen individual
rights. Firstly, the right to informational self-determina-
tion, as an expression of the fundamental right to have and
develop a personality, formally recognised in the article 22
of the Universal Declaration of Human Rights. More spe-
cifically, it would sustain the right to receive information,
guaranteed by the article 10 of the European Convention
on Human Rights. Similarly to what serendipity by design
seeks to achieve, this right sustains political debate, truth
finding, social cohesion, avoidance of censorship, and self-
development (Eskens et al. 2017).
Designing for serendipity might also help to defend what
has been termed as the ‘right not to be deceived’, which
means to be treated with dignity and respect and not to
be manipulated and “hyper-nudged” by means of big data
(Yeung 2017; see also note 12). Of course, this depends
on how designing serendipity is conceived, in particular
whether it is considered not only passive serendipity (type
A) but also active serendipity (type B and type C).
In this context, the right to informational and decisional
privacy and, more generally, data protection law, remain fun-
damental. Not only to increase control over personal data
and profiling (e.g. the principle of intervenability)20 but, for
example, also to enable what Hildebrandt (2017) defines as
‘agonistic machine learning’, that is, demanding companies
or governments that base decisions on machine learning to
‘explore and enable alternative ways of datafying and mod-
elling the same event, person or action’.
20
As an ENISA study (Danezis et  al. 2015) summarizes: “Interve-
nability ensures intervention is possible concerning all ongoing or
planned privacy-relevant data processing, in particular by those per-
sons whose data are processed. The objective of intervenability is the
application of corrective measures and counter-balances where neces-
sary.”
In summary, serendipity is a valuable experience for both
individuals and democratic societies, particularly in digital
environments in which is technically volatile. It is not in
danger but it can be cultivated to blur the boundaries of filter
bubbles and echo chambers. More generally, it can sustain
media pluralism and strengthen individuals’ human rights.
Before to analyse the ethics of serendipity and the challenges
and opportunities as a principle, the main characteristics of
designing for serendipity are briefly explored.
An information architecture for serendipity
Programming for serendipity sounds like an oxymoron. It
is in fact a subjective experience that is only discernible in
hindsight and for which the unexpected plays a fundamental
role, as it is accidental and unpredictable. Serendipity, in
effect, cannot be created ‘on demand’. While natural, ‘pure
serendipity’ is not amenable to generation by a computer
(van Andel 1994), artificial serendipity can actually be cul-
tivated by creating opportunities for it through the design
of physical, digital, and learning environments (Race and
Makri 2016; Melo 2018). So far, however, relatively little
research has been undertaken to assess how well existing
approaches to information interaction support and assess
serendipity.
Even though there is no set of systematic requirements for
a system that facilitates serendipity in digital environments
yet, many scholars agree that the main functional require-
ments are to enable the “anomaly” or the chance encounter
to trigger an event, support the user in “connecting the dots”
so as to make him/her reach a significant surprise outcome
(McCay-Peet and Toms 2017). In general, this eventu-
ally translates into four major goals; on the one hand, to
design serendipity into recommender systems to (1) provide
(pseudo)random content, (2) provide content that intersect
users’ profile, so that a user might discover new ideas and
interests, outside one’s filter bubble, and (3) provide content
that is politically and ethically challenging (that is, diverse
viewpoints and worldviews), eventually resulting in seren-
dipitous encounters (belief change), in order to minimize the
effects of echo chambers. On the other hand, it is intended,
more generally, as (4) to cultivate serendipity by design, i.e.
throughout empowering tools to randomize, control and be
aware of the production and consumption of information.
Serendipity is therefore an affordance, an opportunity that
the environment offers to the user that is able to grasp it,
as an intuitive invitation (Gibson 2014). At the same time,
there are affordances for serendipity itself (Bjorneborn
2017). Clearly, the challenge is how to operationalize
them. From a design perspective, two general approaches
are default rules and active choosing (Sunstein 2017b). In
other words, a user may take both a non-purposive or pas-
sive information behaviour (which generally supports Type
13

A serendipity)  and a purposive or active (mostly Type B
and C) (Kotkov et al. 2016). These will be now analysed in
relation to the information environments that support them.
Serendipity-driven recommender systems
Personalized recommender systems (RSs) have presented
since long the problem of excessive homogeneity. RSs
research in fact began to move beyond accuracy and experi-
menting with serendipity (Ge et al. 2010). Indeed, user sat-
isfaction does not always correlate with very accurate RSs.21
This, however, does not guarantee that mainstream platforms
will take it into consideration. RSs have already been criti-
cized to not sufficiently account for serendipity (Matt et al.
2014).
Optimizing current RSs for more serendipitous recom-
mendations is not a trivial task of course. Promising attempts
have been already done long ago. Campos and Figuereido
(2002) famously pioneered an information retrieval software
that resulted in 52.7% of (pseudo)serendipitous suggestions.
Despite the subjective character of the results, they showed
that explicitly programming for serendipity is possible. Yet
there is still no consensus on the definition of serendipity
in RSs (Kotkov et al. 2016). It is, however, a compound of
other concepts like novelty, utility, diversity, unfamiliarity
and unexpectedness.
Even though design a RS is a multi-criteria decision mak-
ing problem, there is a major theoretical design trade-off
between two main criteria: accuracy and serendipity (Ibid).
Accuracy-driven RSs can lead to more engagement and sat-
isfaction for users (thus revenues for information interme-
diaries) than serendipity-driven RSs. More often, in fact,
serendipitous content may not be liked. Designing for ser-
endipity, however, might increase user satisfaction in the
long-term. Furthermore, seeking and experiencing seren-
dipity is time-consuming. Potentially-serendipitous content
(e.g. a long article or a documentary) may bring the user to
another website, pulling off profitable scrolling. More gen-
erally, it implies a slower consumption of information as it
may require an ‘incubation period’ (McCay-Peet and Toms
2017). Thus, such time may be given away to scrolling, and
so to engagement within the platform. The profit-driven
model of mainstream SNS, for example, tend to prioritize
ephemeral content to durable one, short videos (snippets) to
long ones and snapshots (casual images) to written text as
21
This seems to be true also for consumers’ advertisement satisfac-
tion. In fact, there is no consensus yet on the effectiveness of targeted
advertisement (Zuiderveen Borgesius et  al. 2014). On the contrary,
contextual advertisement may actually increase serendipitous encoun-
ters more than the personalized one. Given the oligopoly of the
advertisement industry, there might indeed be incentives not to meet
demand and supply as efficiently as possible.
13
U. Reviglio
they better lock-in users in their “walled-garden” (Derakh-
shan 2016).
Eventually, other relevant criteria need to be considered
to create serendipity-driven RSs, in particular challenging
ethical and political information. Afterwards, it may be pos-
sible to give users the autonomy to filter information. For
instance, by choosing the degree of serendipity in a slider
that moves between personalization, generalization and
randomization. It is indeed their dynamic relation which
can actually sustain an ever more serendipitous environ-
ment (Carr 2015). It is possible, then, to make chronologi-
cal ordering the default mode for media content delivery to
subtract from any potentially biased curation.
Serendipity by design
There are actually countless ways to empower users, extract
value from their profiles meaningfully, to illustrate blind
spots, increasing diversity, and to stimulate creative asso-
ciations. In many cases, designing for serendipity is indeed
a form of nudging. Yet it is also possible to nudge users
without introducing manipulative measures (Helberger
et al. 2016). Such approach is embraced by Floridi (2016b),
who advocates what he calls ‘pro-ethical design’ or “tol-
erant paternalism” that, in short, aims to modify the level
of abstraction of the choice architecture by educating users
to make their own critical choices and to assume explicit
responsibilities.22
First of all, visualization and discovery tools such as the
use of real-time transparency dashboards help to increase
users’ awareness, autonomy and, as a consequence, the
potential to access and encounter more diverse and, eventu-
ally, serendipitous information (for example, see Nagulendra
and Vassileva 2016). There are indeed several visualization
tools that can increase control and awareness about a user’s
information diet.23
In this context, a significant design choice might be to
afford more profiles per user. The management of multiple
filtering can in fact nudge users to subtract from the deter-
mined path offered by algorithms (Bozdag and Timmermans
22
For instance, Sunstein (2017) proposed that social media like
Facebook could create a “serendipity button” for news and opin-
ions, allowing people to opt in, especially during elections. Similarly,
related stories at the bottom of a “post” seem to help in counteracting
misinformation or simply enriching a user perspective in a serendipi-
tous way (Bode and Vraga 2015).
23
Notably, plug-ins like Balancer and Scoopinion can show a histo-
gram of the user’s liberal and conservative pages or what journals a
user use to read and for how long, with the aim to increase aware-
ness so that they would make their reading behaviour more balanced.
Other tools like Social Fixer and Gobo (a social media aggregator
built by MIT) provide more interactive control over design choices
and information filtering processes.
Serendipity as an emerging design principle of the infosphere: challenges and opportunities
2011; Hildebrandt 2017). Similarly, to afford users to
“empathize” with—namely browsing—the information flow
of other’s users, friends, and even groups may be serendipi-
tous.24 Either, to exploit others’ profile to filter your own
feed (what is called algorithmic recommender personae, see
Harambam et al. 2018). Yet, it must always be acknowledged
the paradox of choice, that is, the more to choose from, the
less that is chosen.
The above design choices and visualization tools repre-
sent only a speculative attempt to frame main affordances
for an architecture for serendipity that could supplement user
experience. Its aim would be to provide users with ‘insti-
tutionalized serendipity’ (Merton and Barber 2006). Such
endeavour sustains certain values that will be now discussed.
A preliminary ethical framework of serendipity
in digital environments
In the last decade applied ethics has developed into an even
more practical discipline. Such “design turn in applied eth-
ics” (Van den Hoven et al. 2017) acknowledges that the
needs and values of users are considered in their own right
and not simply as a side constraint on successful implemen-
tation. This perspective helps to develop pragmatic meth-
odologies and frameworks (like Value Sensitive Design and
pro-ethical design) that help to make moral and ethical val-
ues integral parts of research and development (Friedman
et al. 2002).
To fill the gap between the prescriptions derived from
general theories and the results of the prescriptions in the
world of policy making and the professional practice it is
fundamental to acknowledge that technology is not neutral
with respect to values (Winner 1980). Artefacts can influ-
ence and alter our moral behavior through a reciprocal
process (Verbeek 2011), especially ICTs that are egoipoi-
etic technologies or technology of construction of the self
(Floridi 2011). At the same time, information is a Rawlsian
“primary good”, a good that everybody requires as a con-
dition for well-being (van den Hoven and Rooksby 2008).
Yet without individuals’ capabilities to make sense of it as
to transform information into knowledge, individuals can-
not devise and pursue their rational life plans (Sen 1990).
Also, perfectly rational decision-making is at times neither
practical nor attainable (Simon 1955; Tversky and Kahne-
man 1974).
Fundamental questions to open the debate are thus
addressed; in this context, how and why can serendipity
inform designers, ethicists and, eventually, policy-makers?
24
For example, MIT Lab created for Twitter a plug-in called Flip-
feed which basically provides to the users the possibility to scroll the
feed of a random individual which resides in a far ideological spec-
trum from our own.
What are the values that it sustains in relation to filter bub-
bles and echo chambers? How can these actually help indi-
viduals and democracies to flourish in information societies?
And, also, how to translate serendipity into general norms
and design requirements? There can be a “good measure-
ment” of it? Last but not least, could it be guaranteed a
“desirable degree” of serendipity?
A philosophical perspective
Serendipity is a powerful concept strictly related to other
philosophical ones such as Aristle’s poiesis, Jung’s synchro-
nicity, and Nietzsche’s amor fati (Lupo 2012; Olma 2016). It
is indeed a quite complex value that can be conceptualized in
different ways. In this paper, only an initial effort to outline
a preliminary ethical framework will be done.
From an epistemological perspective, serendipity is an
observation of a surprising fact followed by a correct abduc-
tion (Merton and Barber 2006). In short, abduction—also
called retroduction or inference to the best explanation—as
a premise has a known phenomenon while as a conclusion a
probable explanation. According to Peirce (1992), abduction
is considered the first step of scientific reasoning, arguing
that “all the ideas of science come to it by way of abduc-
tion” (Ibid). Of course, it may also lead to fallible inferences
such as affirming the consequent. Yet these are amendable
hypothesis. Abductive reasoning is indeed considered as a
more powerful mode for navigating the uncertainties of the
information age because it would better support genuine
doubt and skepticism (Cunningham 2001), which is indeed
associated with greater intention to seek information, more
news media use and more frequent online opinion expression
(Yamamoto et al. 2018).
Serendipity is a capability and, as such, can be considered
in terms of virtue ethics. As sagacity and curiosity are often
acknowledged as fundamental virtues, there are two other
primary ones: humility and courage (Lupo 2012). On the one
hand, humiliation is necessary in the face of the imponder-
able occurrence of chance so as to radically question one’s
own beliefs. On the other hand, courage is needed to fol-
low such chance, and to turn it into an opportunity against
its potential risks. It is also necessary for any individual
engaged in the task of getting rid of beliefs and prejudices
that curb the cognitive path. As scientists strive to embrace
such attitude, individuals can also benefit from being nudged
to be sagacious and critical active-seekers alike in online
settings. Critically, to experience serendipity it is needed to
take the risks of failure, frustration and “wasted time”.
Serendipity is also strictly related with randomness,
chance and, therefore, uncertainty and mystery. Offline
activities often imply different information behaviours than
13

online ones.25 In the current social media landscape, any-
thing is designed to be more efficient and most of the choices
are delegated to algorithms. The pursuit for serendipity can
thus be interpreted as a technical attempt to preserve and
support a sense of freedom and mystery that is available in
less networked information environments. It might indeed
become fundamental to maintain in the infosphere the pleas-
ing feelings that elevate accidental discoveries to sensations
of serendipity. In future algorithms, it will be indeed neces-
sary to leave some things open to chance (Domingos 2015).
As serendipity cannot be programmed but only culti-
vated, its design is more probabilistic and, therefore, it
is supposed to imply a wider degree of (pseudo)random
encounters compared to an efficiency-driven information
architecture.26 Seeking serendipity would also account for
what in statistical jargon are called outliers—deviant cases
that do not follow the statistical model and that might com-
promise its predictive power—that represent the personal
“long tail”. Similarly, Abbott (2008) provided a theoreti-
cal basis for concerns about algorithmic filtering, arguing
indeed the necessity to defend randomness from processes
for search and discovery. Because of evolution and the abil-
ity to adapt to environmental changes, human beings are
“antifragile” (Taleb 2012), in the sense that they actually
benefit and get better from shocks, randomness and uncer-
tainties. In other words, extremely accurate algorithms may
reduce human resilience.
Personalization might indeed overly reduce pain and
uncertainty (Gal 2017) so as to create only an illusion of
knowledge, eventually pushing people to radicalize their
25
For example, you can glance at the beginning, the middle and the
end of a book, or a newspaper, so you can find a page by chance, or
a particular paragraph or line. Online one may miss that strange feel-
ing of mystery and accident when we come across one particular line
by chance. It may feel somehow irrationally significant because we
are, indeed, also irrational creatures. This feeling of awe and surprise
is so deeply entrenched in human nature that many cultures reflected
it in a practice called “bibliomancy”, the art of predicting the future
with books (Forsyth 2014). When an ancient Greek wanted to know
the future, he would take a copy of the Iliad and let it fall open at a
random page, point at a line, read it out, and that would be his fate.
This was the Sortes Homericae. The Romans did the same thing with
Virgil—the Sortes Virgilianae. Even medieval chaps did that with the
Bible and called it the Sortes Sanctorum. Interestingly, according to
Walpole—who coined the term serendipity—his particular brand of
discovery was referred to by a certain “Mr. Chute” as a Sortes Walpo-
lianae (de Melo 2018).
26
It is important to stress that a serendipity-driven information fil-
tering would actually increase diversity as long as it remains highly
probabilistic—and even purely random—as is it by definition, so that
serendipitous encounters remain highly unpredictable and rare. A
truly perfect personalized serendipity engine might even decrease the
diversity of information and, as a consequence, most of its beneficial
effects. This issue, however, concerns only Type A serendipity and
particularly what is defined as Kairos (see note 33).
13
U. Reviglio
beliefs due to the easiness to receptively confirm them in
online settings (Lynch 2016). Adding to the ‘Socrates para-
dox’ (I know that I know nothing), serendipity helps to inter-
nalize the significance of the ‘unknown unknown’, what we
don’t know that we don’t know. Thus, it preserves necessary
feelings of uncertainty and mystery. Philosopher of infor-
mation Floridi (2015a) argues that a low and stable degree
of uncertainty leads to increased degrees of liberalism and
toleration, as well as more efficient flows of information.
Similarly, Marcus (2010) acknowledges that the emotional
anxiety provoked by perceptions of conflict, even if unpleas-
ant (e.g. zemblanity, see note 19), is the only means of forc-
ing a rational re-evaluation of one’s currents beliefs.
The value of serendipity should also be considered on
the light of future technological developments. Advances
in so-called artificial intelligence, machine learning and
the semantic web have the potential to enable algorithms to
make ever more sophisticated recommendations. Whereas
virtual reality, augmented reality and the Internet of Things
will definitely blur the anachronistic distinction between
online and offline, giving rise to the vision of Ambient Intel-
ligence,27 or even leading to a ‘social physics’ (Pentland
2015). Serendipity can indeed be designed in physical places
and other emerging contexts.
An educational perspective
Serendipity embraces certain values that can represent future
educational goals and political struggles of the digital era.
Not only its design generally sustains creativity and innova-
tion but its teaching helps to stimulate proactive information
seeking, inter-disciplinary skills and enrich our algorithmic
imaginary. As a consequence, it might help to solve the main
challenges to experience diversity online (see Hoffmann
et al. 2015) and, at the same time, sustain the four modali-
ties of expressing voice28 (see Harambam et al. 2018).
Firstly, since serendipity is intimately related to research
and discovery, to occur it requires an environment that stim-
ulates platform’s transparency and user control, it implies,
therefore, an empowerment of individuals. Contrary to many
legal formulations in data protection which are framed in
terms of harms, and thus freedom from algorithms, serendip-
ity can be considered a positive freedom, as an individual,
27
Ambient Intelligence refers to the eventual future vision in which
automatic smart online and offline environments interact with each
other and take an unprecedented number of decisions for us and about
us in order to cater to our inferred preferences. It may actually repre-
sent a new paradigm in the construction of knowledge (Hildebrandt
and Koops 2010).
28
Building on the work of Hirshman, Harambam et  al. (2018),
intend the concept of voice as the possibility to exert control over the
data-driven processes that shape news provision.
Serendipity as an emerging design principle of the infosphere: challenges and opportunities

social and technical engagement for the discovery of new, Table 1 Summary of the main opposing phenomena and antithetical
unexpected and valuable information. values that the design of serendipity tends to favour and disfavour
When users are treated as active-seekers, then, they can Favours Disfavours
perceive their findings as triumphs of personal agency, intui-
(Pseudo)Randomization Predictability
tion, and inspiration, and as a self-reinforcing expectation,
Chance Efficiency
increase their freedom from algorithms. Maturity of infor-
Noise Signal
mation societies is indeed a matter of people’s expectations,
Autonomy Delegation
not just technological or economic development (Floridi
Information seeking Passive consumption
2016a). If one analyzes the phenomenon of personalization
Uncertainty Security
through the lens of serendipity, it is possible to rebalance
Pluralism Political polarization
the asymmetric power between human agency and computer
Belief change Radicalization
delegation, and between users and corporations’ rights.
Interdisciplinarity Specialism
Secondly, serendipity has a semantic value, in the sense
Divergency Convergency
that it can expand users’ algorithmic imaginary (Bucher
Realism Hedonism
2017). Our current conceptual toolbox indeed is no longer
Weak ties Homophily
fitted to address new ICT-related challenges and, as Floridi
Hard news Soft News
(2015b) argues, this is a risk because ‘the lack of a clear con-
Innovation Redundancy
ceptual grasp of our present time may easily lead to nega-
Education and self-learning Engagement
tive projections about the future: we fear and reject what we
Long-term satisfaction Instant gratification
fail to semanticise’.29 Serendipity can actually represent a
narrative for emerging information societies and enrich the
‘ecology of the self’, as the ethics of self-poiesis (Floridi
2011). For instance, it can strengthen individual’s reactance Limitations and unintended consequences
to bias-driven algorithms. Either, it might help to recognize
one’s own responsibility in the spreadability of serendipitous There are several arguments against serendipity as a design
information, given that users are actually one another’s cura- principle. As many other principles, indeed, serendipity
tors (Race and Makri 2016, p. 5.2) (Table 1). works well in theory while in practice shows many limita-
Finally, managing serendipity can help us to navigate tions. Several challenging issues indeed lie on its definition,
our hyper-complex world more resiliently. Scholars have its normative translation and, eventually, in its measurement.
remarked that the researcher of the future has to be able to First of all, one of the main challenge is indeed the meas-
work in several fields temporarily and to find similarities urement of serendipity, which mainly depends on the sub-
between apparently heterogeneous concepts (Lupo 2012). jectivity of the phenomenon. Individuals have very different
Serendipity, in fact, more often comes out among interdis- thresholds for considering something serendipitous. Seren-
ciplinary scholars (Edward Foster and Ellis 2014) and it dipity is not a purely discrete concept (Makri and Blandford
boosts interdisciplinary research (Darbellay et al. 2014). In 2012). In its assessment, serendipity is a continuum to cover
the informal learning that derives from HCI, a serendipitous the entire spectrum of different degrees of surprise. There-
proclivity represents a powerful soft-skill to tackle the chal- fore, serendipity cannot be considered as a ‘good measure-
lenges of the information era and a new fundamental idea ment’ that is valid, reproducible and accurate, thus it can
that contribute to rethink the current context of education become object of dispute (see Kroes and van de Poel 2015).
(Cobo and Moravec 2011). Indeed, it can help to develop It can be established, nonetheless, a reasonable or justified
abilities of self-learning through accidental learning (Kop consensus on how to operationalize it and measure it. On the
2012; Yadamsuren and Erdelez 2016). After all, the expe- base of which evaluation criteria, then, information archi-
rience of serendipity eventually embraces philosophical, tectures and RSs should be designed remains open-ended.
educational and, thus, political values. It is, then, not only a Measurements of serendipity can be both quantitative and
design principle but also an ethical one. qualitative scales (McCay-Peet and Toms 2017). It is indeed
possible to observe serendipity in controlled research envi-
ronments, throughout surveys or by directly observing infor-
mation encountering behaviour (Kotkov et al. 2016). Quan-
titative measurements may then be based on mixed method
29
approach and other evaluation criteria. Eventually, these
The word ‘semaniticise’ did not really exist and it is defined by
might complement each other to potentially come up with an
Floridi (2011) as the way in which “we make sense of our environ-
ment, of ourselves in it, and of our interactions with and within it.” additional key performance indicator. Nevertheless, a “desir-
(p. 564). able degree” of serendipity remains subjective; in the case

13

of passive serendipity, how much potentially-serendipitous
information should be provided both in the quantity of un/
expected information recommended and the degree of such
un/expectedness?30 Perhaps, somewhat counter-intuitively,
an optimal balance may be found personalizing serendip-
ity, albeit there might be significant privacy implications
(McCay-Peet and Toms 2017).
A related-risk lies in potential manipulations of serendip-
ity (particularly passive serendipity), due to its relatively
conceptual fuzziness and, again, its subjective character. For
instance, Google CEO Eric Schmidt (2006) envisioned a
future where people and technology come together to cre-
ate ‘a serendipity engine’ “where you don’t even have to
type” (2006).31 Instead of create serendipity and offer it ‘on
a plate’, however, it is more sustainable to create oppor-
tunities for it by supporting strategies that may increase
its likelihood, thus empowering users to create their own
personal ‘recipes’ (Makri et al. 2014). Mere simulations of
serendipity, in fact, may produce vague, washy attempts to
make the world a better place (Olma 2016). One risk is to
legitimize whatever recommended content with serendipity
and to muddle it with sensationalism, propaganda or even
disinformation and misinformation (so-called fake news).
For example, one of the principles of captology—Kairos—
refers to the ability to provide the right content, at the right
person in the right moment (Fogg et al. 2002). More often,
this is merely fake, manipulative serendipity.32
From a regulatory perspective, optimize social media for
serendipity does not appear promising. Above all, it might be
considered as a form of stealthy paternalism naturally suit-
able to handling. Yet inequality would persist. Challenge-
averse individuals who would ideally benefit more from an
architecture for serendipity, are naturally those less willing
30
In this case, the un/expectedness is referred to the probability
assigned by the engineers that a certain information may be liked.
The challenge is indeed how to balance the probability distribution of
the information prioritized.
31
For example, Google implicitly afford users to experience such
passive serendipity with the search button “I am Feeling Lucky”.
However, it is rarely used. Instead, a more constructive design choice
Google implemented was a toggle feature directly into the central
search bar that allowed to generalize the query results (See Carson
2015). This option, however, has been recently relegated into the
advanced options, no more evident in the Google results page.
32
Kairos is a Greek divinity, personification of the “opportune
moment”. It might be conceived as the “evil brother” of serendipity.
It is in fact a surprising event yet not necessarily serendipitous. It is
the epitome of convergent systems. Also, the individual is persuaded
and has no autonomy whatsoever. Yet these hyper-personalized and
persuasive techniques will be increasingly employed. Once tasted,
however, such algorithmic recommendations may be impossible to
live without. To resist it, then, it may be beneficial to employ con-
vergent systems also for serendipitous recommendations. Afterwards,
the boundary between Kairos and Serendipity is only how ‘good’ the
intentions (and skills) of designers and engineers are.
13
U. Reviglio
to play with it. On the contrary, those already prone to it
are those more willing to benefit from it. Given the het-
erogeneity of users, to avoid further inequality a regulation
may be required. As said, the actual loss for platforms might
be users’ engagement and, consequently, it would threaten
their revenues. Therefore, it might even represent a political
challenge.
Still, the beneficial effects of an architecture for serendip-
ity are uncertain. Looking for serendipity is often time-con-
suming, risk-taking, sometimes distracting. Being actively
exposed to challenging information—eventually serendipi-
tous—may not necessarily result in experience of diversity,
de-polarization and/or more tolerance. In some cases, the
opposite is true. Polarized individuals radicalize further
while exposed to challenging information (so-called backfire
effect) (Quattrociocchi et al. 2016). Any intervention should
eventually be measured accordingly. Experiencing serendip-
ity especially requires an attitude of open-mindedness that
is, above all, an educational issue.
Therefore, while from a theoretical perspective serendip-
ity can be advocated as a beneficial design principle, from
a pragmatic perspective it is a highly problematic endeav-
our. Not only there are still potential improvements in the
understanding of serendipity from a theoretical perspective
but there is also significant design, psychological and edu-
cational research gap. Yet, we argued that serendipity has
to be conceived both passive and active. In other words, an
architecture for serendipity should balance personalization,
generalization and randomization, and may be beneficial
as insofar it is both implicit and explicit, convergent and
divergent.
Conclusions
This paper analyzed the phenomenon, the discussion and
the design of serendipity in digital environments, particu-
larly in relation to online personalized media content. Far
from having systematically framed its boundaries, the paper
constitutes a preliminary ethical analysis of serendipity in
digital environments. Serendipity is indeed a fundamental
experience of social reality, and a complex process which
occurs in all realms of human life, most notably in science.
It embraces fundamental values such as curiosity, research,
discovery, progress and identity development. Even if ser-
endipity is often an implicit design goal, in some cases other
values may be privileged instead. Serendipity is not only
a by-product of the information environment but it can be
limited and facilitated by design choices and recommender
systems. Therefore, it should be taken into consideration in
the design process. In practical terms, this means to provide
also content that is less likely to be appreciated by a user.
Serendipity as an emerging design principle of the infosphere: challenges and opportunities
Thus, the main implication of designing for serendipity is
the control (or the halting) of hyper-personalization.
Given that personalization of media content and nudging
techniques will become increasingly sophisticated over time,
the pursuit for serendipity can actually help to cultivate a
healthier infosphere. It is in fact possible to design, cultivate
and experiment information architectures that can mitigate
the risks of redundant and/or persuasive personalization by
exposing users to more potentially meaningful information.
This is especially relevant for political news consumption
in social network sites and video sharing platforms. In this
sense, designing for serendipity should involve democratic
principles such as being exposed to challenging and alter-
natives perspectives. As a consequence, it would help to
balance personalization, generalization and even (pseudo)
randomization.
Furthermore, serendipity can help to enrich our con-
ception of media diversity as well as strengthen individual
rights. Indeed, it encompasses fundamental phases of pro-
duction and consumption of information: from information
filtering to information behaviour. Its design requires content
diversification and users’ interactive control of information.
Yet, if not conceived broadly—taking into account both
passive and active serendipity—it can even legitimize per-
suasion by design. Serendipity is not only a form of nudg-
ing but also a capability. As such, it implies a design that
does not exploit “inferred weaknesses” —passive consump-
tion, hedonism or anything that might trigger  addictive
loops—but ideally empowers individuals to incrementally
discover novel, diverse and meaningful information. Thus,
it nudges users to be open-minded and critically proactive in
seeking and managing information. Thus, explicit personali-
zation is particularly strengthened.
Finally, methodological limitations and potential unin-
tended consequences have been analyzed. Interdisciplinary
discussions and, eventually, consensus on the definition, the
design and the assessment of serendipity need to be further
discussed. As of now, many research gap remain. Also, the
practical effectiveness in bursting filter bubbles and soften
echo chambers is questionable. Digital literacy and critical
thinking remain fundamental preconditions for a beneficial
expression of serendipity in digital environments. Moreo-
ver, under the current business model of mainstream social
media, its flourishing is actually profoundly limited. Even if
huge challenges as a design principle persist in reality, as an
ethical horizon serendipity can fruitfully inform and inspire
designers, educators and users.
Acknowledgements This research is funded by the ERASMUS MUN-
DUS program in Law, Science and Technology (LAST-JD) coordinated
by University of Bologna.
Compliance with ethical standards
Conflict of interest The author declares that he has no conflict of inter-
est.
Informed consent Research for this paper did not involve animal or
human participants; neither was there a need to request informed con-
sent from anyone.
References
Abbott, A. (2008). The traditional future: A computational theory of
library research. College & Research Libraries, 69(6), 524–545.
Ahmadi, M., & Wohn, D. Y. (2018). The antecedents of incidental
news exposure on social media. Social Media + Society, 4(2),
2056305118772827.
Albanie, S., Shakespeare, H., & Gunter, T. (2017). Unknowable
manipulators: Social network curator algorithms. arXiv preprint
arXiv:1701.04895.
Auray, N. (2007). Folksonomy: The new way to serendipity. Interna-
tional Journal of Digital Economics, 65, 67–88.
Backstrom, L. (2013). News feed FYI: A window into news feed (p. 6).
Menlo Park: Facebook for Business.
Björneborn, L. (2017). Three key affordances for serendipity: Toward
a framework connecting environmental and personal factors in
serendipitous encounters. Journal of Documentation, 73(5),
1053–1081.
Bode, L., & Vraga, E. K. (2015). In related news, that was wrong: The
correction of misinformation through related stories functional-
ity in social media. Journal of Communication, 65(4), 619–638.
Bodo, B., Helberger, N., Irion, K., Zuiderveen Borgesius, F., Moller, J.,
van de Velde, B., … de Vreese, C. (2017). Tackling the algorith-
mic control crisis—The technical, legal, and ethical challenges
of research into algorithmic agents. Yale JL & Tech., 19, 133.
Bogers, T., & Björneborn, L. (2013). Micro-serendipity: Meaningful
coincidences in everyday life shared on Twitter. iConference,
2013, 196–208.
Bozdag, E., & Timmermans, E. (2011). Values in the filter bubble
Ethics of Personalization Algorithms in Cloud Computing. In
Proceedings, 1st International Workshop on Values in Design–
Building Bridges between RE, HCI and Ethics, Lisbon.
Bozdag, E., & van den Hoven, J. (2015). Breaking the filter bubble:
Democracy and design. Ethics and Information Technology,
17(4), 249–265.
Bucher, T. (2017). The algorithmic imaginary: Exploring the ordinary
affects of Facebook algorithms. Information, Communication &
Society, 20(1), 30–44.
Campanario, J. M. (1996). Using citation classics to study the inci-
dence of serendipity in scientific discovery. Scieontometrics,
37(1), 3–24.
Campos, J., & Figueiredo, A. D. (2002). Programming for serendipity.
In Proceedings of the AAAI fall symposium on chance discov-
ery—The discovery and management of chance events.
Carr, N. (2016). Utopia is creepy: And other provocations. New York:
W W Norton & Co Inc.
Carr, P. L. (2015). Serendipity in the stacks: Libraries, information
architecture, and the problems of accidental discovery. College
& Research Libraries, 76, 831–842.
Carson, A. B. (2015). Public discourse in the age of personalization:
Psychological explanations and political implications of search
engine bias and the filter bubble. Journal of Science Policy &
Governance, 7(1).
13

Cobo, C., & Moravec, J. (2011). Invisible Learning. Toward a New
Ecology of Education. Col·lecció Transmedia XXI. Laboratori
de Mitjans Interactius/Publicacions i Edicions de la Universitat
de Barcelona. Barcelona.
Colton, S., & Wiggins, G. A. (2012). Computational creativity: The
final frontier?. In Ecai (Vol. 2012, pp. 16–21).
Cunningham, D. J. (2001). Fear and loathing in the information age.
Cybernetics & Human Knowing, 8(4), 64–74.
Danezis, G., Domingo-Ferrer, J., Hansen, M., Hoepman, J. H., Metayer,
D. L., Tirtea, R., & Schiffner, S. (2015). Privacy and data pro-
tection by design-from policy to engineering. arXiv preprint
arXiv:1501.03726.
Darbellay, F., Moody, Z., Sedooka, A., & Steffen, G. (2014). Interdis-
ciplinary research boosted by serendipity. Creativity Research
Journal, 26(1), 1–10.
de Melo, R. M. C. (2018). On serendipity in the digital medium:
Towards a framework for valuable unpredictability in interac-
tion Design.
De Rond, M. (2014). The structure of serendipity. Culture and Organi-
zation, 20(5), 342–358.
Delacroix, S. (2018). Taking turing by surprise? Designing digi-
tal computers for morally-loaded contexts. arXiv preprint
arXiv:1803.04548.
Derakhshan, H. (2016). Social Media Is Killing Discourse Because
It’s Too Much Like TV in MIT Technology Review. Accessed
Jan 15, 2018, from https://www.technologyreview.com/s/602981/
social-media-is-killing-discourse-because-its-too-much-like-tv/.
DeVito, M. A. (2017). From editors to algorithms: A values-based
approach to understanding story selection in the Facebook news
feed. Digital Journalism, 5(6), 753–773.
Domingos, P. (2015). The master algorithm: How the quest for the
ultimate learning machine will remake our world. New York:
Basic Books.
Dunbar, K., & Fugelsang, J. (2005). Scientific thinking and reasoning.
The Cambridge Handbook of Thinking and Reasoning, 705–725.
Dylko, I., Dolgov, I., Hoffman, W., Eckhart, N., Molina, M., & Aaziz,
O. (2018). Impact of customizability technology on political
polarization. Journal of Information Technology & Politics,
15(1), 19–33.
Edward Foster, A., & Ellis, D. (2014). Serendipity and its study. Jour-
nal of Documentation, 70(6), 1015–1038.
Erdelez, S. (1997). Information encountering: a conceptual frame-
work for accidental information discovery. In Proceedings of
an international conference on Information seeking in context
(pp. 412–421). Taylor Graham Publishing, London.
Erdelez, S. (2004). Investigation of information encountering in the
controlled research environment. Information Processing &
Management, 40(6), 1013–1025.
Erdelez, S., & Jahnke, I. (2018). Personalized systems and illusion of
serendipity: A sociotechnical lens. In Workshop of WEPIR 2018.
Eskens, S., Helberger, N., & Moeller, J. (2017). Challenged by news
personalisation: Five perspectives on the right to receive infor-
mation. Journal of Media Law, 9(2), 259–284.
Floridi, L. (2011). The informational nature of personal identity. Minds
and Machines, 21(4), 549.
Floridi, L. (2015a). The politics of uncertainty. Philosophy & Technol-
ogy, 28(1), 1–4.
Floridi, L. (2015b). The onlife manifesto. Cham: Springer.
Floridi, L. (2016a). Mature information societies—A matter of expec-
tations. Philosophy & Technology, 29(1), 1–4.
Floridi, L. (2016b). Tolerant paternalism: Pro-ethical design as a reso-
lution of the dilemma of toleration. Science and Engineering
Ethics, 22(6), 1669–1688.
Fogg, B. J., Lee, E., & Marshall, J. (2002). Interactive technology and
persuasion. The Handbook of Persuasion: Theory and Practice.
Thousand Oaks: Sage.
13
U. Reviglio
Forsyth, M. (2014). The unknown unknown: Bookshops and the delight
of not getting what you wanted. London: Icon Books Ltd.
Friedman, B., Kahn, P., & Borning, A. (2002). Value sensitive design:
Theory and methods. University of Washington technical report,
pp. 02–12.
Gal, M. S. (2017). Algorithmic challenges to autonomous choice.
Michigan Telecommunications and Technology Law Review,
2017.
Ge, M., Delgado-Battenfeld, C., & Jannach, D. (2010). Beyond accu-
racy: Evaluating recommender systems by coverage and seren-
dipity. In Proceedings of the fourth ACM conference on Recom-
mender systems (pp. 257–260). ACM, New York.
Gibson, J. J. (2014). The ecological approach to visual perception:
Classic edition. Hove: Psychology Press.
Gillespie, T. (2014). The relevance of algorithms. Media technologies:
Essays on communication, materiality, and society, p. 167.
Granovetter, M. S. (1977). The strength of weak ties. In Social net-
works (pp. 347–367). Chicago: University of Chicago Press
Gup, T. (1997). Technology and the end of serendipity. The Chronicle
of Higher Education, 44(21), A52.
Harambam, J., Helberger, N., & van Hoboken, J. (2018). Democratiz-
ing algorithmic news recommenders: How to materialize voice
in a technologically saturated media ecosystem. Philosophical
Transactions A, 376(2133), 20180088.
Helberger, N. (2011). Diversity by design. Journal of Information
Policy, 1, 441–469.
Helberger, N., Karppinen, K., & D’Acunto, L. (2016). Exposure diver-
sity as a design principle for recommender systems. Information,
Communication & Society, 21, 1–17.
Hendler, J., & Hugill, A. (2013). The syzygy surfer:(Ab) using the
semantic web to inspire creativity. International Journal of Crea-
tive Computing, 1(1), 20–34.
Hildebrandt, M. (2009). Profiling and AmI. In The future of identity in
the information society (pp. 273–310). Berlin: Springer.
Hildebrandt, M. (2017). Privacy as protection of the incomputable self:
Agonistic machine learning.
Hildebrandt, M., & Koops, B. J. (2010). The challenges of ambient
law and legal protection in the profiling era. The Modern Law
Review, 73(3), 428–460.
Hoffmann, C. P., Lutz, C., Meckel, M., & Ranzini, G. (2015). Diver-
sity by choice: Applying a social cognitive perspective to the
role of public service media in the digital age. International
Journal of Communication, 9(1), 1360–1381.
Hoven, J. van den, Miller, S., & Pogge, T. (Eds.). (2017). Designing
in ethics. Cambridge: Cambridge University Press.
Karppinen, K. (2008). Media and the paradoxes of pluralism. The
Media and Social Theory, 27–42.
Keymolen, E. (2016). Trust on the line: A philosophycal exploration
of trust in the networked era.
Kop, R. (2012). The unexpected connection: Serendipity and human
mediation in networked learning. Journal of Educational Tech-
nology & Society, 15(2), 2–11.
Kotkov, D., Wang, S., & Veijalainen, J. (2016). A survey of seren-
dipity in recommender systems. Knowledge-Based Systems,
111, 180–192.
Kroes, P., & van de Poel, I. (2015). Design for values and the defini-
tion, specification, and operationalization of values. Handbook
of Ethics, Values, and Technological Design: Sources, Theory,
Values and Application Domains, 151–178.
Krotoski, A. (2011). Digital serendipity: Be careful what you don’t
wish for in The Guardian International Edition. Accessed Jan
15, 2018, from https://www.theguardian.com/technology/2011/
aug/21/google-serendipity-profiling-aleks-krotoski.
Loepp, B., Hussein, T., & Ziegler, J. (2014). Choice-based preference
elicitation for collaborative filtering recommender systems. In
Serendipity as an emerging design principle of the infosphere: challenges and opportunities
Proceedings of the SIGCHI Conference on Human Factors in
Computing Systems (New York, NY, 2014), pp. 3085–3094.
Lupo, L. (2012). Filosofia della serendipity (Vol. 73). Napoli: Guida
Editori.
Lutz, C., Hoffmann, P. C., & Meckel, M. (2017). Online serendip-
ity: A contextual differentiation of antecedents and outcomes.
Journal of the Association for Information Science and Tech-
nology, 68(7), 1698–1710.
Lynch, M. P. (2016). The Internet of us: Knowing more and under-
standing less in the age of big data. New York: WW Norton
& Company.
Makri, S. (2014). Serendipity is not Bullshit. Paper presented at the
EuroHCIR 2014, The 4th European Symposium on Human-
Computer Interaction and Information Retrieval, 13 Sep 2014,
London, UK.
Makri, S., & Blandford, A. (2012). Coming across information seren-
dipitously—Part 1, p. A process model. Journal of Documenta-
tion, 68(5), 684–705.
Makri, S., Blandford, A., Woods, M., Sharples, S., & Maxwell, D.
(2014). “Making my own luck”: Serendipity strategies and how
to support them in digital information environments. Journal
of the Association for Information Science and Technology,
65(11), 2179–2194.
Maloney, A., & Conrad, L. Y. (2016). Expecting the unexpected: Ser-
endipity, discovery, and the scholarly research process (white
paper), Thousand Oaks: SAGE.
Marcus, G. E. (2010). Sentimental citizen: Emotion in democratic
politics. University Park: Penn State Press.
Matt, C., Benlian, A., Hess, T., & Weiß, C. (2014). Escaping from
the Filter Bubble? The Effects of Novelty and Serendipity on
Users’ Evaluations of Online Recommendations. In Proceed-
ings of the 35th International Conference on Information Sys-
tems (ICIS2014), Auckland, New Zealand.
McCay-Peet, L., & Toms, E. G. (2013). Proposed facets of a seren-
dipitous digital environment. In Teoksessa iConference 2013
Proceedings, ss. 688–691.
McCay-Peet, L., & Toms, E. G. (2017). Researching serendipity in
digital information environments. Synthesis Lectures on Infor-
mation Concepts, Retrieval, and Services, 9(6), i–i91.
Meckel, M. (2011). “Sos—save our serendipity”, Personal Blog.
https://www.miriammeckel.de/2011/10/11/sos-save-our-seren
dipity/.
Merton, R. K., & Barber, E. (2006). The travels and adventures of
serendipity: A study in sociological semantics and the sociol-
ogy of science. Princeton: Princeton University Press.
Nagulendra, S., & Vassileva, J. (2016). Providing awareness, expla-
nation and control of personalized filtering in a social network-
ing site. Information Systems Frontiers, 18(1), 145–158.
Negroponte, N. (1996). Being digital. New York: Vintage.
Olma, S. (2016). In Defence of Serendipity. Watkins Media Limited,
2016.
Pariser, E. (2011). The filter bubble: How the new personalized
web is changing what we read and how we think. New York:
Penguin.
Pasquale, F. (2015). The black box society: The secret algorithms that
control money and information. Harvard: Harvard University
Press.
Peirce, C. S. (1992). The essential Peirce: Selected philosophical writ-
ings (Vol. 2). Indiana: Indiana University Press.
Pentland, A. (2015). Social physics: How social networks can make us
smarter. New York: Penguin.
Pinch, T. J., & Bijker, W. E. (1984). The social construction of facts and
artefacts: Or how the sociology of science and the sociology of
technology might benefit each other. Social Studies of Science,
14(3), 399–441.
Polly, J. A. (1992). Surfing the internet. An Introduction. Wilson
Library Bulletin, 66(10), 38.
Powers, E. (2017). My news feed is filtered? Awareness of news per-
sonalization among college students. Digital Journalism, 5(10),
1315–1335.
Quattrociocchi, W., Scala, A., & Sunstein, C. R. (2016). Echo cham-
bers on Facebook. Available at SSRN: https ://ssrn.com/abstr
act=2795110.
Race, T., & Makri, S. (2016). Accidental information discovery: Cul-
tivating serendipity in the digital age. Cambridge: Chandos
Publishing.
Reviglio, U. (2017). Serendipity by design? How to turn from diver-
sity exposure to diversity experience to face filter bubbles in
social media. In International Conference on Internet Science
(pp. 281–300). Springer, Cham.
Rubin, V. L., Burkell, J., & Quan-Haase, A. (2011). Facets of serendip-
ity in everyday chance encounters: A grounded theory approach
to blog analysis. Information Research, 16(3), 27
Schmidt, E. (2006). How we’re doing and where we’re going. Google
Inc. Press Day 2006.
Schönbach, K. (2007). ‘The own in the foreign’: Reliable surprise-
an important function of the media? Media, Culture & Society,
29(2), 344–353.
Sen, A. (1990). Justice: Means versus freedoms. Philosophy & Public
Affairs, 19(2), 111–121.
Sen, A. (2005). Human rights and capabilities. Journal of Human
Development, 6(2), 151–166.
Shannon, C. (1948). A mathematical theory of communication. Bell
System Technical Journal, 27, 379–423.
Shearer, E., & Gottfried, J. (2017). News use across social media plat-
forms 2017. Pew Research Center, Journalism and Media.
Simon, H. A. (1955). A behavioral model of rational choice. The Quar-
terly Journal of Economics, 69(1), 99–118.
Smith, A. (2018). Many Facebook users don’t understand how the site’s
news feed works. Pew Research Center, Journalism and Media.
Stiegler, B. (2017). The new conflict of the faculties and functions:
Quasi-causality and serendipity in the anthropocene. Qui Parle,
26(1), 79–99.
Sunstein, C. R. (2009). Going to extremes: How like minds unite and
divide. Oxford: Oxford University Press.
Sunstein, C. R. (2017a). # Republic: Divided Democracy in the Age of
Social Media. Princeton: Princeton University Press.
Sunstein, C. R. (2017b). Default rules are better than active choosing
(Often). Trends in Cognitive Sciences, 21(8), 600–606.
Sunstein, C. R. (2017c). In praise of serendipity in The Economist,
Accessed Feb 4, 2018, from https://www.economist.com/news/
books-and-arts/21718464-social-media-should-encourage-chanc
e-encounters-not-customised-experiences-praise.
Taleb, N. N. (2012). Antifragile: Things that gain from disorder
(Vol. 3). New York: Random House.
Thurman, N. (2011). Making ‘The Daily Me’: Technology, economics
and habit in the mainstream assimilation of personalized news.
Journalism, 12(4), 395–415.
Thurman, N., & Schifferes, S. (2012). The future of personalization at
news websites: Lessons from a longitudinal study. Journalism
Studies, 13(5–6), 775–790.
Turing, A. M. (1950). ‘Computing machinery and intelligence’. Mind,
59, 433–460.
Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty:
Heuristics and biases. Science, 185(4157), 1124–1131.
van Andel, P. (1994). Anatomy of the unsought finding. serendipity:
Origin, history, domains, traditions, appearances, patterns and
programmability. The British Journal for the Philosophy of Sci-
ence, 45(2), 631–648.
van den Hoven, J., & Rooksby, E. (2008). Distributive justice and
the value of information: A (broadly) Rawlsian approach.
13

Information Technology and Moral Philosophy, p. 376. Cam-
bridge: Cambridge University Press
Verbeek, P. P. (2011). Moralizing technology: Understanding and
designing the morality of things. Chicago: University of Chi-
cago Press.
Vicario, M., Bessi, A., Zollo, F., Petroni, F., Scala, A., Caldarelli, G.,
& Quattrociocchi, W. (2016). The spreading of misinformation
online. Proceedings of the National Academy of Sciences, 113(3),
554–559.
Winner, L. (1980). Do artifacts have politics? Daedalus, 109(1),
121–136.
Yadamsuren, B., & Erdelez, S. (2016). Incidental exposure to online
news. Synthesis Lectures on Information Concepts, Retrieval,
and Services, 8(5), i–i73.
Yamamoto, M., Hmielowski, J., Beam, M., & Hutchens, M. (2018).
Skepticism as a political orientation factor: A moderated media-
tion model of online opinion expression. Journal of Information
Technology & Politics, 15(2), 178–192.
13
U. Reviglio
Yaqub, O. (2018). Serendipity: Towards a taxonomy and a theory.
Research Policy, 47(1), 169–179.
Yeung, K. (2017). ‘Hypernudge’: Big Data as a mode of regulation by
design. Information, Communication & Society, 20(1), 118–136.
Zarsky, T. Z. (2002). Mine your own business: Making the case for the
implications of the data mining of personal information in the
forum of public opinion. Yale JL & Tech, 5, 1.
Zuckerman, E. (2013). Rewire: Digital cosmopolitans in the age of
connection. New York: W. W. Norton & Company.
Zuiderveen Borgesius, F. J., Trilling, D., Moeller, J., Bodó, B., De
Vreese, C. H., & Helberger, N. (2016). Should we worry about
filter bubbles? Internet Policy Review. Journal on Internet Regu-
lation, 5(1), 16.
Publisher’s Note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.