The Presentation of Self Online

Introduction

Web users are engaging in computer-mediated self-expression in varying ways. The technology that enables this is developing fast, and how it does so is influenced by many more factors than just the needs of the people it touches. This chapter explores literature about the impact of computer-mediated self-expression on:

  • people's everyday lives.
  • individual self-expression and exploration.
  • interactions and relations with others.
  • interactions with and expectations of society and community.

We ground our discussion in established literature about non-digital self-expression and identity from the social sciences. This raises the key theme of individuals desiring control over how others see us, yet wanting to behave in a way that is authentic, or consistent with their internal identities. There is also emphasis on the collaborative and collective nature of identity formation; that is, our self-presentation fluctuates depending on the people we're with, the situation we're in, and norms of the society we're part of. The focus on face-to-face interactions and embodiment leads us to draw contrasts between online and offline experiences, and to look at the substitutes for the body in digital spaces.

The extent to which online and offline identities interact and overlap is hotly debated. Is creating an online identity a chance to reset, to reshape yourself as an ideal? Or are you simply using it to convey true information about what is happening in your daily offline life? Is it a shallow, picture of you, or a forum for deep self-exploration? How does the way one portrays oneself in digital spaces feedback to ones offline self-presentation? We explore these questions in section 3.

Section 4 examines social media and blog use, including how one's imagined audience affects self-presentation in public, and how context collapse might occur when the actual audience is different to expected. There are several examples of techniques for managing who sees which 'version' of oneself, and the types of 'versions' of self that are commonly seen to be constructed on social media, and with what degree of transparency they are linked together. Most of the longitudinal studies in this space are of teenagers and young people, who have never known a world without social media, and who may incorporate it naturally and seamlessly into their daily practices, thus making it a core part of their identity during formative years. I draw a contrast between the relationship-driven architecture of contemporary social networking sites, and the more personal, customisable blogging platforms which preceded them. Studies of bloggers and blogging communities reveal some different priorities and habits than what is common practice today, and offer insight into how online self-presentation is evolving.

Throughout literature from both social and computer sciences, privacy is a common concern. In section 5 we look further at how tensions between users and the privacy settings of systems they use impact on personal information disclosure. Does self-censorship affect identity formation? How do people weigh up the risks and benefits of exposing themselves online? This is particularly pertinent for future systems development, as more and more people become aware of state surveillance, for-profit data collection, and their diminished rights over their personal data.

Finally we introduce the relatively new Web Science concept of Social Machines in section 6 in order to recapture the circular interdependencies between humans, technologies, and communities. We propose to build on current work of describing and classifying social machines to better account for the individual perspectives of participants.

Ultimately we posit that online is simultaneously a reflection, a distortion, an enhancement, and a diminishment of the offline world. They impact each other in complex ways, particularly with regards to self-presentation and identity formation. The various theories and studies described in this chapter form the basis for which we conduct the investigative and technical work in the remainder of this thesis.

My perspective on this review

I'd like to take a moment to note that whilst reading various studies about young peoples' reactions to and interactions with rapidly evolving digital technologies from the 2000s, it occurred to me that the subject of these studies is in fact my own age group. Some of the results are instinctively familiar to me; I was there, I experienced these things. Some are ridiculous. I don't know how my first-hand experience of growing up with technology (I was born in the same year as the Web, and my parents were early adopters) affects my reading of these studies, or my ability to study others' use of technology, but it is something I ponder.

Performing the self

The obvious place to start when embarking on a discussion about self-presentation is Goffman [goffman1959]. In The Presentation of Self in Everyday Life, Goffman posits several, now well-established, theories using drama as a metaphor:

  • Everyone is performing. The front-stage of our performance is what we create for others - the audience - to see, so that they may evaluate and interact appropriately with us.
  • We also have a back-stage; how we act when there is no audience, or an audience of our team. Our team participate alongside and collude with us on the front-stage.
  • Our performances have both conscious and unconscious aspects. That is, we consciously give information about ourselves to others in order to manage their impression of us, but we also unconsciously give off information that others may pick up on and take into account when deciding how to interact with us.
  • Both actors and audiences are complicit in maintaining the cohesion of a situation. Performances break down if actors break character, deliberately or accidentally, or if there is a mismatch between parties' definition of the situation.

These theories emphasise the collaborative or social nature of self-presentation, and apply to face-to-face interaction.

Whilst Goffman's dramaturgy refers mostly to body language, a related theory is Brunswik's lens model [brunswick56, lens01], part of which suggests that individuals infer things about others based on "generated artifacts", or things left behind. In [bedrooms02] this model is used to study how personal spaces (offices and bedrooms) affect observers' assessments of the characteristics of the owner of the space. This study links individuals to their environments by:

  • self-directed identity claims (eg. purposeful decorations like posters or use of colour);
  • other-directed identity claims (eg. decorations which communicate shared values that others would recognise);
  • interior behavioural residue (ie. "physical traces of activities conducted within an environment");
  • exterior behavioural residue (ie. traces of activities conducted outside of the immediate environment which nonetheless provide some cues as to the personality of the environment occupant).

Self-presentation is largely unconscious in the physical realm and comes naturally the most people. People may also use in-crowd markers (like a shirt with a band logo on) consciously to send certain messages to people who will recognise them, whilst not drawing any attention from people who won't [boydfacid]. Later in this chapter I look at how our presence in digital spaces fail and succeed to take the place of the physical body when it comes to interactions and identity formation.

The self in context

By reflexively adjusting one’s perception of self in reaction to society, people construct their individual identity. [boydfacid]

Development of personal identity is not only something that happens internally. We are strongly influenced by feedback (conscious and unconscious) from others around us, as well as the particular setting and culture in which we find ourselves. How we react to things outside of our control in part determines our identity construction, and some people adjust their behaviour in response to feedback more than others [snyder74]. Thus identity is socially constructed, and often is dynamically adjusted according to context [boydfacid].

The project of the self

Giddens [giddens84] looks at the relationship between macro and micro views of the world, acknowledging that broader effects of society impact individual behaviour, and vice versa, with neither one being the primary driving force. This suits well my ideas about online self-presentation, confirming the complex interplay between technological affordances, individual actions, and the place of both in a cultural and social context.

Giddens argues that self-identity is an aggregation of a person's experiences, an ongoing account, and a continuous integration of events. In contrast to Goffman's dramaturgy, Giddens downplays the role of an audience, and in contrast to Brunswik's lens theory, he downplays what we can learn from the traces someone leaves behind. Giddens argues that self-identity cannot be uncovered from a moment, but something which is ongoing, over time. Modern society, according to Giddens, affords us more freedom to create our own narratives to determine our self-identity. In the past, rigid social expectations dictated our roles for us. However, increased choices about what to do with ourselves may also increase stress and prove problematic. Awareness of the body is central to awareness of the self, as the body is directly involved in moments we experience in daily life. As we are now explicitly constructing a narrative about our identities, rather than having one ascribed to us by society, the self is an ongoing project which takes work to maintain [giddens91].

The focus on explicit actions and decision making about self-presentation is pertinent when it comes to digital representations of identity.

Extending the self

Early to mid 20th century philosophers and social scientists complicate notions of the 'self' by combining it and extending it with our physical surroundings, and this view emerged long before the Web. Heidegger expresses technology as coming into being through use by a human; when tools are used the tool and its user do not exist as independent entities, but as the experience of the task at hand (using the example of a carpenter hammering, unaware of himself or his hammer) [manhammer]. McLuhan discusses media, literate and electronic, from the printing press and electric light to radio, TV and telephone, and its impact on how we communicate. He places communication technologies as simultaneously extensions of and amputations of our bodies and senses, which continuously and fundamentally re-shape the way we (humans) see and place ourselves the world [mcluhan]. More recently, Clark's Extended Mind Theory uses the example of a notebook as a means of externally processing information that would otherwise be carried out by the brain, drawing the external world in as party to our cognitive processes [clarkmind].

The next logical step is to consider how the modern digital technologies of Web and social networking can also be considered extensions of the self, and this is addressed in part by Luppicini's notion of Technoself [technoself]. Technoself incorporates (amongst other things) extension of the self through physical technology embedded in the body (cyborgology); in our changing understanding of what it is to be, as life is extended and augmented through advancing healthcare; but also in our relationships with our virtual selves. This is not a topic into which I will dive deeply from a philosophical standpoint, but the idea of the Web and online social networks as extensions to the self rather than as separate entities or concepts is worth bearing in mind as this thesis proceeds to explore the complexities of intertwined digital and offline identities.

Offline to online… and back again

When people use digital technologies to communicate, they are passing a version of themselves through the filter of the platform they use. In this section I discuss the relationships between online and offline selves.

One might assume that the disembodied nature of the interactions lets people have more control over how they are "seen" by others; Turkle claims that it is "easy" to create and tweak a perfect self online [Turkle2015]. In reality this control is affected by a great number of factors. Turkle writes that computer-mediated communication is predictable even in the ways it is unpredictable, and that people seek out this predictability in preference to face-to-face interactions, and in preference to facing emotional vulnerability. Turkle's argument assumes that we have full understanding and control of the digital systems we use and the audiences we are reaching through them. It presumes we know exactly how and where and when the data we input will be output in the short and long term, and how others will interpret it.

Turkle's argument neglects that at every crossroad in these 'predictable' systems are stationed unpredictable humans, perhaps with conflicting interests and motivations, from the conception of a social system, to its realisation and use. Turkle has been studying for decades how people explore, experiment and find themselves through technology, and her overriding narrative is of a desire to express an idealised version of the self; one that is not subject to any interpretation other than what the expresser desires. However, as discussed further in the next section numerous studies of social media users find a variety of other types of motivation for participation.

Keen [digitalvertigo] on the other hand emphasises the risks that individuals become trapped by technology of which they have neither understanding nor control. Being swept up in cultural technology trends, social media users may unwittingly become "prisoners" of a carefully curated digital "hyperreality", the importance of which supplants their offline lives. Similarly, [inreallife] suggests that "fantasy gets in the way of real progress" when it comes to self improvement, but studies only examples of "catfishing", where individuals create exaggerated online profiles in order to deliberately mislead potential romantic interests. The idea that the online self replaces rather than supplements the offline self also misses the nuances of how and why people use social media in the first place. Nonetheless, I agree that online worlds are certainly not just a mirror of the offline.

On the contrary, not only are online identities some form of reflection of offline identities, but the inverse can be true as well. In ethnographic studies of teenagers' use of mobile apps for socialising, it is reported that crafted online personas both reflect and feed back into teenagers' true sense of self [appgen]. Numerous studies from the field of psychology, reported in [infinitereality], demonstrate various ways in which playing a role online or in a virtual world reflects back and directly changes people's offline behaviour. Avatars can be used to build confidence and reveal suppressed personality traits. Relatedly, study participants who were asked to interact from behind avatars which conveyed different appearances of age, race, and body type expressed affinity with these previously unfamiliar experiences, and responded differently in personality evaluations before and after. A longitudinal study of teenage girls roleplaying online, a process through which they developed their identity through narrative, revealed positive impacts on their confidence, through new friendships [youth].

There is a long history of assessments of online interactions which proclaim that the self-centered nature of social media makes people narcissistic, that competition for reputation isolates us [digitalvertigo], and that the construction of an ideal self or facade is damaging, especially to young people. However, [appgen] argues that narcissism is not created by the Web, but enabled by it, as an existing need for validation is more readily satisfied.

Several studies [insta17, Rousseau17, haferkamp10] find that people's self confidence or body image drops after viewing the online profiles of people who appear to be more attractive or more successful than them. [wellbeing17] explains that passive use of social media is what appears to have a negative impact on people, whereas active use has a positive effect on well-being. This is shown in several studies, including [toma10], which explores the beneficial effects of browsing one's own Facebook profile; [nemer15] which discusses how selfies can empower marginalised communities; and [noland06] which finds a positive impact on self-esteem of teenaged girls who engage in "auto-photography".

It is also worth bearing in mind that experience in the virtual world can cause physical reactions - laughter, tears - and the virtual and the physical blur together in the subject stream of experiences, adding to an identity which is made of virtual and physical events blended together [youth].

Authenticity and integrity

In an interview, Facebook founder Zuckerberg said that "having two identities for yourself is an example of lack of integrity" [fbeffect]. This received public backlash at the time, and on several more occasions as Facebook and other social networking sites imposed real name policies, sometimes linked to an official ID [boyd2012politics]. This removes a level of identity control which many people take for granted. A particularly clear example can be seen in the reaction of a community of drag queens who were used to being able to interact online using chosen names. Real name requirements removes a vital aspect of fluidity from their interactions, glossing over the seams which provide poignant performance material and a route towards a group identity [lingel2015face, barmann14drag]. Additionally the requirement for real names and the ability to report 'fake names' became a mechanism whereby a marginalised group could be harassed and silenced [hotmess2015selfies]. In general, an attitude that people should be happy to connect all of their identities together under a single legal name is an expression of social privilege: a result of having no features or proclivities which are socially censored.

Sandberg, whilst COO of Facebook in 2012, commented that profiles as detailed self portraits is a "shift towards authenticity" [digitalvertigo], but her organisation's notion of 'integrity' as a single complete version of oneself that is the same no matter to whom one is presenting is somewhat at odds with authenticity. Neither are people "intraviduals", caught between competing identities as claimed by Conley in 2009 [conleyintra], but expressing aspects of themselves appropriately and according to context. This is behaviour which we have already established via Goffman as ordinary offline, and so shouldn't be considered unusual online. One's 'authentic' self-presentation may be partial or moderated and no less genuine for that. Indeed, some individuals find they are more able to express their authentic selves online than they are offline due to oppression or social expectations which are disjoint from their core values.

However, in highly commercialised or competitive online environments, 'authenticity' is a quality to strive for, to maintain an audience. A Web search for 'authenticity on social media' will reveal a plethora of guides on how to craft an 'authentic' persona, how to maintain personal-but-not-too-personal ties with one's audience so that they see that you're just like them. For individuals who set out to explore and express their identity online, this can be a tricky world to navigate [duffy15b]. True authenticity in online communities is seen as disjoint with self-promotion and celebrity; popularity implies a reduction in authenticity, perhaps linked to 'selling out' or 'pandering' to an audience [donathboyd04, ellison07]. Whereas authenticity is often seen from an outside perspective as always something manufactured, an idealized reality [Marwick2010].

The idea that online spaces are under control of their owners suggests others may be suspicious of their authenticity [tong08] but reinforcement of social identity from others can counteract this. Warranting theory describes how information that appears to be outside of the subject's control—for example, a message posted publicly by a friend on someone's profile—can reinforce the trustworthiness of the other profile data to an outside observer [warrant, warrant1, warrant2].

Dishonesty and deception

"Some argue that distinct contexts are unnecessary and only encourage people to be deceptive. This is the crux of the belief that only those with something to hide need privacy." - [boydnp10]

Most people like to consider themselves to be quite honest in their communications with friends, family and acquaintances. However, even honest people routinely modulate what they share, omitting and sometimes falsifying information in order to reduce social friction, avoid confrontation, defuse awkward situations, or to save face [buller1996interpersonal, burgoon1989maintaining]. Hancock et. al. introduced the term butler lies to refer to a common use of simple lies to manage communications, such as smoothly exiting from an unwanted conversation [hancock2009butler]. Online, the notion of who our 'friends' are has become increasingly blurred and difficult to define. In such settings, people commonly navigate different social spaces, projecting and varying self-presentation according to the ways they want to be perceived by each [marwick2010TweetHonestly].

Whilst part of tailoring one's presentation to an audience is the ability to carry out some level of deception, with personal communications, there is an implicit expectation of authenticity [aoki2005Ambiguity]. However, online, the need to navigate multiple and uncertain audiences means that we may constantly vary our self-presentation. Authenticity becomes a social construct derived from the social context and how we wish to be perceived by a given audience [boyd2002faceted]. We may be deceiving, at least to some extent, nearly constantly without even being conscious of it.

Deception has long been studied, both within and outwith the HCI community. Traditionally, deception has been cast in a negative light [bok1978lying], to be used only if no other option is available. In the 1980s, however, communications researchers began to investigate the positive aspects of lying, in particular white lies - socially acceptable lies which cause little or no harm to the recipient [camden1984WhiteLies].

In 1992, McCornack cast deception as an understandable response to complexity: "[r]esearchers studying deception recently have begun to argue that deceptiveness is a message property that reflects a kind of functional adaptation to the demands of complex communication situations" [mccornack1009InformationManipulation]. People then manipulate the information which they share as a necessary part of participation in society. This has led to recent work on the positive aspects of deception in human computer interaction, in particular how butler lies are used to ease social situations [hancock2009butler], and how systems can deceive their users for beneficial reasons [adar2013Benevolent].

Several different taxonomies of lying and deceptive behaviours have been proposed [camden1984WhiteLies, depaulo1996lying, lindskold1983categories]; Anolli et al. examined a family of deceptive miscommunications, including self-deception and white lies [anolli2001DeceptiveMiscommunication]. They look at omission of relevant information, concealment using diversionary information, falsifaction and masking with alternative, false information. Of particular interest is their claim that "a deceptive miscommunication theory should be included in a general framework capable of explaining the default communication", that is that deception should not be seen as a psychologically different activity than 'normal' communication. This tallies with the earlier approach of McCornack [mccornack1009InformationManipulation] who situates deceptive messages within the spectrum of information manipulation. This, combined with the lens of Gricean conversational maxims, allows for an explanation of deceptions where some of the truth is told, but information which the speaker knows is relevant to the listener is omitted or obscured [grice1970logic].

Motivations for lying have also been extensively studied in social psychology. Turner et al.'s taxonomy included saving face; guiding social interaction; avoiding tension or conflict; affecting interpersonal relationships; and achieving interpersonal power [turner1975information]. Camden et. al. [camden1984WhiteLies] develop a detailed categorisation of lies to do with basic needs, managing affiliation with others, self-esteem and miscellaneous practices such as humour and exaggeration.

Many malicious or undesirable behaviours are facilitated by the ability to create and alter identities. Astroturfing [cho2011astroturfing] has become common online [zhang2014astroturfing], with corporations and governments employing sophisticated identity management software to carry out large scale operations. Possibly the most famous of this is the 50 Cent Party, hired by the government of the People's Republic of China to post favourable comments towards party policy [fiftycent]. On a smaller scale, sock-puppets — multiple accounts controlled by a single person — are used to skew ideas of consensus and distort discussion in online societies, leading to attempts to automatically identify such accounts [bu2013sock, solorio2013case]. Personas can be constructed for the purpose of trolling, whether it is overtly offensive in order to cause outrage or more subtle manipulation to trick people into wasting effort or taking caricatured positions, and correlations have been shown between enjoyment of trolling and everyday sadism [buckels2014trolls].

Many of these activities are a form of obfuscation, in some way hiding the truth, polluting the data pool and diminishing trust. The ethical issues here are complex and contextual, with the viewpoints of different actors having considerable divergence [brunton2011vernacular].

Another strand of research borrows from information warfare, to look at the possibilities for disinformation. Disinformation tactics are most useful when a channel of information cannot be completely closed, but can be rendered useless by being filled with incorrect, but plausible, assertions in order to lower its overall signal-to-noise ratio [wiki:disinformation]. The intended target of the lie may not be the official recipient of the message: lies can be directed at those who are eavesdropping on the communications channel or surveilling the participants [alexander2010Disinformation]. Techniques used include redaction to remove parts of the message, airbrushing to blur parts of the message and blending to make the message similar to other plausible messages, as well as other forms of information distortion [alexander2010Disinformation].

In chapter 3 I carry out two studies which aim to bring together these general theories of deceptive behaviour with a closer look at how and why people might engage in them online.

Networked publics

Social media technologies blur the boundaries between private and public, and this affects identity performance. Ten years ago, boyd hoped that educators and technologists would succeed in easing the cultural transition for young people into the networked era [boyd-sns07]. She describes social networking sites as a type of "networked publics", technologically-mediated spaces where people can virtually go to interact with their friends, and where they may be subject to observation or interjection by passers-by. Differently to offline public spaces, online publics may be persistent, scalable, searchable, replicable, and/or have invisible audiences. These features of networked publics affect how people express themselves and interact, however they do not directly dictate participants' behaviour [boydnp10]. Networked publics are not only spaces, but collections of people or "imagined community"; different publics can serve different purposes, but can also intersect with each other [boyd2014s].

Over subsequent years, boyd and many others proceed to explore the effects of these differences on those who engage with online social media to different degrees. In this section I recount some of these studies and findings.

A benefit of participation in networked publics is that a wider variety of communities are accessible than offline. Niche identities don't have to be set aside to fit in [appgen]. Online interactions are "not simply a dialogue between two interlocutors, but a performance of social connection before a broader audience" [boyd2014s]. boyd looks specifically at teenagers in networked publics, who she says have sought online spaces in recent years as they are not allowed to 'hang out' any more in physical spaces like malls [boyd2014s, marwickdrama14].

First I reflect on the digital substitutions for the physical body in online social interactions. Then, in comparing and contrasting 'old school' style blogging with contemporary (circa 2013-2017) social networking sites I look deeper into how differences in technological affordances impact peoples' interactions and self-expression.

Audiences for identity performance as well as the context in which the performance takes place are critical, but online both of these may be unknown or dynamic, or both. I'll introduce work around imagined audience and context collapse, both of which pioneer our understanding of identity behaviours in networked publics. When audiences and contexts are known, we can examine how people connect with others and form communities; in the final section I look at trust, social reinforcement of identity, and studies of what social media participants choose to disclose or conceal.

Profiles and embodiment

In Faceted Id/entity [boydfacid], boyd highlights several differences between self-representation offline compared to online. Embodiment is a key factor in self-presentation and she claims that there is considerable difference between performing one's identity through appearance, eg. fashion and body language, when walking into a room, compared to explicitly describing oneself by entering attributes and other personal information into an online form to create a profile. The disembodied nature of online interactions means that people must find new ways to express themselves, and manage the impressions other people have of them, or "a new type of body" [youth]. A lack of control over one's online self presentation is compounded by the inability to visualise - or perhaps even be aware of - the data that is collected by the systems we use. Online activities are logged over time to an extent that most individuals are not aware of; these activities, an individual's expressions given off (Goffman), are used, largely unknowingly, for the commercial benefits of third parties; this constitutes a kind of implicit or unconscious profile. boyd suggests that visualising all of one's personal information that is available online, as well as visualising one's 'audience' or social network connections, would provide an individual with better awareness of, and so better control over, their online image. In boyd's prototype interfaces, users are explicitly asked for personal data in order to build a profile of themselves, and boyd does point out the problematic nature of this, compared with the unconscious or implicit identity performance one conducts in offline social settings.

Counts in [counts09] explores the impact of profile attribute selection on self presentation, and finds that upon completing the values for 10 attributes, participants converge on their "ideal" representation of themselves. This study also finds that free-form attributes are better than ones with preset choices for participants' satisfaction with how they have portrayed themselves. This study does not take into account that most online profiles are created in a particular context, with a particular purpose in mind. Asking participants to express their ideal self-presentation 'in general' vastly oversimplifies reality. Participants are not told who the consumers of the profile they are creating are expected to be, or how it is to be presented; nor are participants given an opportunity to indicate who their expected audience is or what they think the profile is for.

Since boyd's prototypes were designed, social media gained widespread popularity. Most, if not all, mainstream systems request input of explicit personal data to build an initial profile, despite the discord of this activity compared with offline identity expression. However, unlike in boyd's prototypes, it doesn't stop there. Such systems encourage ongoing engagement through adding and messaging contacts with various degrees of publicness, creating status updates to broadcast a current situation, production of creative media content, and feedback on content and updates created by others. As we will see in chapter 3, contemporary social media builds one's profile from various combinations of these online activities, and typically use far more than the explicit data entered by the user to generate a representation of a person. This increases the likelihood that individuals may not have an accurate impression of what this representation looks like to others.

Recent studies confirm that visuals are a key part of expressing identity online. Many focus on selfies as a modern substitute for the body [vanhouse11, lasen15, senft15, frosch15], but [bunnies17] examines self-presentation through other kinds of photos. Examples include humerous images from popular culture or photographs of other things with an overlayed caption, coupled with a tag (eg. #currentstate) that indicates the poster relates to this concept; as well as photos of items that people carry with them day to day. In [papatwit12], self-expression is performed through use of Twitter hashtags, and [food15] suggests that food photography is a means of self-presentation.

Pointing at something and saying that one has chosen it as self-representative makes the assemblage of tags, text, and image a culturally intelligible self-representation [bunnies17].

A lack of embodiment can also have a distinct advantage. In [stendal12] several studies of people with disabilities who use online social systems are reviewed, and reveal findings about increased control over disclosure of disability (which may not be possible offline) and reduced isolation when people are able to interact online.

Next we look more closely at the behaviour of users of Social Network Sites, of which "profiles" are a key feature [boydsns07].

Social Network Sites

In 2007 boyd and Ellison defined Social Network Sites (SNS) to be Web-based, bounded, public or semi-public, and afford creating and viewing connections with or between other users [boydsns07]. They note that users of these systems tend to connect with others with whom they already have a 'real life' social relationship, and present a fairly thorough history of SNS from 1997 onwards, which I won't recount here. This definition is pertinent to this thesis due to its emphasis on profiles, implying self-presentation, as a core feature of SNS. In 2013 they updated their definition to incorporate different types of content and data into profiles; to de-emphasise the traversal of connections (as this became more important to machines than humans); and to emphasise participants' interaction with streams of user-generated content [ellison13]. I will proceed to focus on case studies and experiments which were carried out since these definitions, and due to the rapid pace of change in this area, prioritise those from the last four years.

Many studies of identity formation on social media focus on young people and teenagers. One reason is because this is a crucial point in life for understanding oneself and asserting a personal identity. Other studies approach teenagers as somewhat alien "digital natives", born into a world of social technology which is expected to fundamentally change how they interact with the world compared with older generations, who don't or can't distinguish between online and offline [borndigi]. I will relate the results of these studies, but note that I disagree with the notion of a "digital native" because being born in a particular year or even raised around modern technology does not automatically give one a natural instinct for identity expression in digital spaces, and not even necessarily more opportunity to experiment and reflect than older SNS users [boyd2014s].

Digital communication technologies can help or hinder identity formation. The App Generation [appgen] provides a balanced argument between the pros and cons of teenagers socialising through mobile applications. They find that some applications provide a "prepackaged identity" for users to adopt rather than encouraging experimentation. The affordances of applications shape the forms of expression that are available, and so identity formation is in a way controlled by the application designer. Born Digital [borndigi] suggests that teenagers experiment with identity online, but aren't fully aware, or don't care about, the traces that are left behind when they do so.

In some cases, for example fan communities, self-presentation shifts between a more playful fictional identity performance, and an identity which is closer to 'real life' [baker09]. In others, such as professional self-presentation, individuals lean on automatically generated metrics by the system they use to convey a positive image, with gamification or commodification of the self becoming commonplace [academicqs16]. When SNS provide a platform for professionalising passions such as content creation, [duffy15] notes that participants may be even more vulnerable to the consequences of performing and maintaining one's self-presentation in an exposed online space, as well as the "labour of visibility" that goes into it.

Most people occupy multiple roles offline, find ways to establish and maintain boundaries between them, and continue to do so to different degrees when taking representations of these roles to online spaces. SNS increase the permeability of boundaries, but users employ various tactics to manage their identity when a one-identity-per-person model is imposed on them [quinn15].

[singh15] describes how Twitter users subvert features of the system to express themselves in new ways, as well as reflecting on how changes to the functionality of they system affect how people use it. This supports [papatwit12] which, through content analysis of trending hashtags, also describes how people work around technical constraints of Twitter to meet their self-presentation needs.

Even in the early days of SNS (specifically Facebook) beginning to rise to popularity when use was overflowing into the workplace, the access control settings offered by Facebook were considered too complicated to enable most people to realistically manage connections with both professional and personal contacts from a single profile, despite the potential advantages of connecting with colleagues through the platform [DiMicco07]. More recently, we see that SNS users manage tensions between their multiple roles and the affordances of systems by segregating their audience across using multiple platforms. The interview study in [Zhao16] found that sharing decisions across multiple sites are made primarily based on the known audiences of the different sites, and the content being shared. This study also recounts previous work on motivations for using different SNS, including to connect with old friends, and share pictures, which feed into decisions taking regarding content sharing. A similar study found that family was a crucial audience to whom more private sharing was desired [Farnham11], and findings in [Vitak14] indicate that Facebook users desire to re-asssert their offline boundaries when online, and concurs that managing this through the tools that Facebook provides is cost-intensive. Facebook itself compounds this issue by using identity information as a "social lubricant" which encourages people to make new connections [ellison11fb].

Blogging and personal homepages

Personal homepages and blogs have been around for considerably longer than SNS, though remain a comparatively specialist practice. It is widely accepted that blog or website owners have more control over their online space than do users of SNS, [intlblog, markus06], including freedom to innovate with the site's appearance and thus explore more individualistic aspects of the online self [alist05]. Relatedly, communities of bloggers are not owned or controlled by a single entity [dennen09].

Through observations of over 200 blogs within a particular community and semi-structured interviews with 40 bloggers, [dennen09] identifies five aspects which affect how bloggers build their identity: name and blog title; descriptive attributes; post content; voice; affiliations; and visual design. All of these are subject to change over time, and sometimes major offline transitions can cause a shift to a new pseudonym or blog altogether; often the audience is invited along however. Blogs are often designed to reinforce community norms, to enhance a sense of belonging; as a result, the community develops and evolves its own identity, which in turn influences how newcomers choose to present themselves. Bloggers' contributions are fragmented across different domains, and where their writing style and topics constitute a part of their identity, so it fragments their identity.

Blogging communities are traditionally more accepting of pseudonyms but [dennen09] notes that distrust is not of other community members, but rather of personally-known community outsiders who may accidentally stumble across blog entries.

Earlier in this chapter I mentioned Brunswik's lens model which describes a way in which identity can be constituted through physical traces left behind. This model has also been used to understand how observers make personality judgments about people based on the traces left in their digital space, ie. personal homepages [gosling08, markus06, vazire04, papacharissi02].

The importance of themes and designs of blogs and homepages is emphasised by [dennen09], who mentions that whilst some blog consumers use a feed reader to receive new content from the blogs they are interested in, they often click through to the original post to view it in the context of the author's own space.

On the other hand, [blogdesign] takes a snapshot of a random sample of blogs in 2003 and maps the state of the blogosphere through analysing visual elements in depth and tracking commonalities. The conclusion is that significant customisation of blog templates was in fact relatively rare, with most people only slightly tweaking colours or adding custom images. A likely explanation for this is that bloggers lacked the technical expertise to do so.

Studies of blogging communities outside of the US demonstrate that blogging is not a uniform practice that can be understood as a whole [intlblog]. Certain communities (in this case, Muslim ones) which are seen by outsiders as homogeneous use blogging to highlight their uniqueness and individuality. Others (for example in China) emphasise their ethnicity and culture as a key part of their identity. Blogs from the Paris Banlieues in fact had a direct impact on how the mainstream media portrayed their plight; an example of how personal identity expression in networked publics was able to affect a broader social understanding of that identity. The overriding message from these studies is cultural taboos and offline societal context affect narrative about identity, and this is reflected online.

Imagined audience

The audience to whom one performs is critical in forming the context in which one is performing [boydnp10]. On SNS, people are often expressing themselves to multiple audiences simultaneously. When people are aware of this, they take different strategies when it comes to navigating what they share; individuals with many followers on Twitter practiced self-censorship (only posting things they are happy for the worst-case audience to read) and practicing coded communication (strategically targeting some posts at some audience members, and others at others, to maintain overall interest) [Marwick2010].

However, given the many possible ways in which Tweets can be discovered and consumed ([Marwick2010] questioned people who post publicly) it is virtually impossible to determine the actual audience for one's content. Thus, people imagine who their audience is likely to be, and express themselves accordingly. Obviously these imaginings, which may stem from understanding of the affordances of a particular platform, or a particular community or topic of discussion, impact how people express themselves online.

[litt12] theorises about how the imagined audience is synthesised, and draws in Giddens' structurational framework, noting a combination of macro- (social roles, technical affordances) and micro-level (individual motivations, technical skills) factors. [litt12] ultimately concludes that asking people about their imagined audience is prone to errors or misinterpretations, as imagined audience is a concept which is both difficult to measure and difficult to express.

Relatedly, as people perform in networked publics, they must contend with a "networked audience," who are not connected only with the performer, but also with each other [Marwick2010].

Context collapse

I have so far discussed how people attempt to map boundaries from their offline lives into their online interaction spaces, and the notion of imagined audience. Context collapse occurs when boundaries come down and personas intended for different audiences are merged [Marwick2010]. The consequences of this may range from slight social awkwardness, to direct breaches of privacy and potential danger, and have been examined in a variety of different circumstances, such as [daviscontext, contextwesch, duguay14, duguay13].

Thanks to the properties of networked publics such as searchability and persistence, contexts may also collapse when information is consumed later, or through a different systems, whereby it may be interpreted differently by the consumer than how it was originally intended [boydnp10].

As we look forward to how SNS and online self-presentation in general will evolve, we must consider how the lines people have drawn around their contexts are tethered to particular (versions of) systems. What happens when these systems change, merge, or disappear? As designers of new systems, we must be cognisant of the role technical affordances play in creating, enabling, and destroying social boundaries.

Everybody knows I’m a dog

"While once viewed as a set of technologies built in resistance to the ugliness of the dot-com era, social media is now intertwined with neoliberal capitalism and data surveillance" - [boyd15]

We are rapidly moving into a world where information about nearly every aspect of our lives is becoming sensed, recorded, captured and made available in digital form. Data is captured and shared voluntarily, as tools invite ever more intimate participatory surveillance [albrechtslund2008Participatory]. While the abundance of information traces has unlocked a wide range of new kinds of applications (eg. [Akker:2014:TRP:2684563.2684638, Consolvo:2008:ASW:1357054.1357335]), the creation and potential for disclosure poses new threats to individual privacy and autonomy. The overall lack of transparency by manufacturers regarding how they are capturing and handling personal information has created a heightened sense of unease among many, in addition to the potential threats dealing with their unintentional disclosure or misuse [fife2012privacy, metzger2004privacy, featherman2010reducing].

Various data and surveillance scandals involving private companies and governments [dcent, crit12] that gained media attention mean that awareness of surveillance and personal data collection is growing amongst the general public. There are many studies examining peoples' awareness of and attitudes towards privacy and surveillance on SNS, but I will not detail them here. During studies in the 1990s, Westin defined three categories to describe how ordinary people feel about privacy: "pragmatists", "fundamentalists" and "unconcerned." Privacy pragmatists accept that there may be tradeoffs between benefits to information sharing and the intrusiveness of requests for information. Fundamentalists distrust organisations which request personal data. The unconcerned are comfortable with sharing personal data with organisations in exchange for services [krane2002privacy]. Westin found, prior to the Web becoming mainstream, that approximately half of the general public are pragmatists; just over half of the remainder are fundamentalists, and a minority are unconcerned. Suffice it to say that people do care about privacy, and are just finding new ways to manage it [boydnp10], contrary to what certain tech executives might claim [zuckpriv, googpriv]. My main concern in terms of this thesis is how privacy infringement might impact presentation of self. We have already seen that online identity performance may alter a general understanding of oneself, which reflects in the offline world. So I must ask: when people self-censor online due to privacy concerns, how does this stifle self-expression, and in turn impact internal identity construction?

Furthermore, implications of our online sharing decisions affect more than just ourselves; "interpreted selves" are created by recognising patterns across millions of people [boydpriv12].

There is often an asymmetry about the collection and use of data. To take a relatively prosaic example, Facebook introduced 'read receipts' on messaging, which indicate when a user has seen a message. This feature has been shown to cause anxiety when present in email systems, as users seek to maintain their responsiveness image, the impression which they project to others about how they respond to input and partition their attention [tyler2003can]. Once ambiguity about attention has been removed, a whole class of white lies - 'The internet was bad, I couldn't check my messages' - are no longer possible, and people develop alternative strategies, such as not opening messages until they feel prepared to respond. The key difference in the context of SNS is that the user does not have the same degree of control over the channel - email receipts can be switched off, but SNS offer different levels of control.

The social aspects of privacy relate to what DeCew terms expressive privacy - a freedom from peer pressure and an ability to express one's own identity [decew1997pursuit]. Nissenbaum's contextual integrity [nissenbaum2004privacy, nissenbaum2009privacy] seeks to understand "appropriate sharing", looking at the ways in which flows of information are governed by norms, which may be easily violated as technological systems repurpose and share data.

The commoditised self

Social systems which involve content creation (like YouTube) or knowledge generation (like Wikipedia) are commonly seen as cooperative communities, whose participants generate value both for each other and also for the organisation behind the system. In [vand09], van Dijck et. al. contest several uncritical manifestos for the business and communal interests of revolutionary Web 2.0 peer-production. They point out that seemingly open co-creation platforms are still profit-driven commercial entities. These entities do not provide tools out of benevolence, but in order to harvest metadata about their users, which they can process and resell. The balance of power between individuals and corporations is not swinging back towards the individual, as proponents of user-generated content sites claim, but the illusion is created that it is.

Users of systems often have little understanding of how their activities are being exploited - or nudged [vand09b]. Even as users are empowered by technology to create media, products, or services they desire, [vand09b] calls into question their agency when participants are being used and manipulated by commercial entities under the guise of community formation or participatory culture. [dbeer08] describes SNS profiles as "commodities, both produced and consumed." In more recent years, awareness of this fact has spread. A popular refrain from advocates of less commercial alternatives is that "if you're not paying for it, you're the product."1, 2, 3

In a similar vein to the previous section, we must also wonder about the impact of external commercial and economic forces which shape the tools and systems people are using to express themselves online. As a contrast, in the second half of this thesis, I focus on decentralised systems, which are potentially much less likely to exploit user metadata for profit.

The ghost in the (social) machine

Social Machines are systems for which the human and computational aspects are equally critical. In most cases, humans do the creative work whilst machines do the administrative tasks [Timbl1999]. Up to now, systems have not been designed to be Social Machines; rather, the concept and definition of a Social Machine is derived from observations of existing (usually Web-based) systems. Many are products of the contemporary social web, on many different scales and in many different domains, often evolving, responding to technological and social developments, and interacting with each other [Hendler2010, ByrneEvans2013, DeRoure2013, Strohmaier2013]. Social Machines can be identified within and across social media networks, within and across online communities, and within and across technological spaces.

Social Machines are pertinent to our work here as they provide a lens through which we can examine sociotechnical phenomena which emphasises the interdependence between humans and technology. In studies of social networks discussed previously in this chapter, humans are considered as users of systems, and discussions focus around how people react to technology, how people behave in the context of particular technical or social constraints, or how people's lives are changed in response to their interactions with and through digital environments. It is important to also reflect upon the ways in which technology evolves or is reconceptualised as a result of passive (mis)use and active (mis)appropriation by humans. We similarly must recognise technical systems in the wider context of society, and include in our dialogue the developers who design and build technical systems, the organisations and legal entities which finance and drive them, and the cultural and economic climate in which they are situated.

Social Machines which have been studied so far have been described and categorised in terms of purpose [DeRoure2013], motivations and incentives, technology used, goals and processes, quality assessment of outputs, and user participation and interaction [Smart2014]. The emphasis in this work is on discussing Social Machines in collective terms; that is, 'a' Social Machine - wherever its boundaries happen to have been drawn (so far these boundaries are typically drawn around the edge of a "service" [Smart2014]) - is considered as a coherent whole. The circumstances of the individual human participants not been given extensive consideration. For any given Social Machine, individual participants are diverse and participate in different ways, with varying goals, motivations and outcomes. As we have seen through studies of social media mentioned previously, they manipulate their online presence(s) so that they may behave in different ways according to different contexts, or may work together to construct a single image controlled by multiple people [Dalton2013]. Such behaviour impacts our understanding of roles, autonomy and awareness, incentives and attribution, and accountability and trustworthiness of participants. Overlooking unique individual perspectives when observing a Social Machine as a whole can cause incorrect assumptions, for example: believing that participants who lie about who they are have negative intentions in a Social Machine whose overall "purpose" is to strengthen social ties. We must also bear this in mind when designing systems, so that a system may grow in response to unexpected actions of participants rather than hampering their explorations.

I argue that due to the complex nature of online identity, understanding nuanced individual behaviours of participants in a more granular way is crucial for Social Machine observation. I advance this argument in the next chapter through an empirical study of a Social Machine centered around creative media production.

Conclusions

Present day social media has dramatically increased participation in publishing and sharing online content. Easy-to-use services lower the barrier to entry for connecting with and pushing thoughts out to an audience. Identities expressed through social media are inherently collaborative; every interaction is pushed to a network, and part of a dynamic cycle of consumption and creation feedback. Commenting on someone else's post automatically links the post to your own profile, and often it appears there as well, accessible from two different contexts. Yet SNS permit little customisation, providing preset options for content or reaction templates, and consistent inflexible designs for profile pages. This tips the balance away from the individual aspect of identity construction performed by bloggers in the earlier years of the Web. Yet blogs and personal homepages are left wanting for a dynamically constructed and low barrier to entry network, which impedes the collaborative aspects of identity construction.

People manage shortcomings with the affordances of both blogging platforms and SNS in different ways—contending with invisible audiences and collapsed contexts, as well as reduced expectations of privacy—through carefully crafted personas, strategically omitting or amending the information they post online, or simply using different platforms for different purposes. We see that there are a multitude of factors which affect people's presentation of self online, which vary according to broader cultural or technological contexts, as well as personal motivations and abilities.

Along with their updated definition of SNS mentioned earlier, [ellison13] calls for social media scholars who are studying individuals or communities online to systematically describe the technology in which their participants are situated, and the practices of the users. Technologies are changing rapidly still, so studies which are a snapshot in time can be linked to a broader discourse or overview, and remain relevant as time progresses, if they situate themselves appropriately. One way of accessing this bigger picture is through an organising framework that helps to record the background state of the environment being studied, and so surfaces connections between work that is otherwise perhaps not directly comparable [vand09b]. In chapter 3 I use several studies of my own on diverse identity behaviours across various social network sites in order to propose such a framework.