Sculpting digital voids: The politics of forgetting on Facebook

Introduction

In 2014, a widely publicized incident brought attention to the proliferation and impact of digital memory features. In his blog ‘Inadvertent Cruel Algorithms’ (2014a), Eric A Meyer, a Web design consultant, writes that earlier in the year, his 6-year-old daughter had tragically passed away due to brain cancer. In the blog, he went on to explain that while he was still in mourning over his daughter’s death, he checked Facebook one day only to find a picture of his daughter featured on the front of his suggested ‘Year in Review’ video montage, surrounded by partygoers and the tag line ‘Eric, here’s what your year looked like!’. Apart from being an unfortunate and untimely ‘coincident’, Meyer (2014a) called this a jarring example of the ‘inadvertent cruel algorithms’ at work on Facebook, stating that:

Algorithms are essentially thoughtless. They model certain decision flows, but once you run them, no more thought occurs. To call a person ‘thoughtless’ is usually considered a slight, or an outright insult; and yet, we unleash so many literally thoughtless processes on our users, on our lives, on ourselves.

For Taina Bucher (2017), the story of Eric Meyer highlights the tensions that inhere in the relations between algorithmic and human decision-making. Echoing Lyotard’s sentiments in The Inhuman, Bucher claims that since computer algorithms and humans ‘think’ in categorically different ways – algorithms being unable to think and discern human complexities – it is questionable whether they should judge humans and their memory practices at all (38). Yet, more than revealing tensions of agency, the story also indicates an emerging way to algorithmically evaluate the relevance (Gillespie, 2014: 175) of people’s past data on Facebook: one which navigates the carefully crafted visibility and invisibility of people’s memory objects.

I want to examine Eric Meyer’s case on Year in Review and its conceptual implications in further detail. The reason is that it importantly demonstrates not only the effects of digital archiving and algorithmic implementations on remembering but also on forgetting. In a context in which everything can resurface and be remembered, forgetting plays an active role in detaching pain from that which remains. Of course, photos of deceased family members and ex-partners are not to be equated with memories of these people. Drawing on Annette Kuhn’s (1995) work, these photographs, videos and posts that have been uploaded on social media platforms are better conceptualized as ‘memory objects’: objects that set the stage for various acts of remembering as well as various engagements with and reactions to these media objects. As traces of memory, these photographs evoke and set the stage for remembering, yet cannot be reduced to memories as such. For Kuhn, remembering is not seen as something given or inevitable but rather as a dynamic act of coming to terms with one’s understanding of the past. It is a performative action, one which is constantly modified and (re-)negotiated.

Examining how people use social media platforms to remember their pasts is, therefore, crucial. But to understand how performative acts of remembering are instantiated and framed on social medial platforms, it is equally crucial to examine the structural dynamics that set the stage for recollection. Using the concept of ‘sculpting digital voids’, I argue that there is a politics of invisibility and forgetting on Facebook’s memory feature, Year in Review. Moreover, I want to use the idea of sculpting digital voids as a conceptual tool to examine the nature and conceptual implications of a set of practices, which aim to render certain memory objects invisible or non-resurfaceable. By incorporating algorithms into its app’s functionality to classify and ensure that certain memory objects do not resurface, Facebook attempts to sculpt digital voids on their memory features. As such, the platform assumes an active role in attempting to prevent certain memory objects from resurfacing on Year in Review and On This Day.

By ‘sculpting’, I mean that digital voids are actively constructed. It is not merely that past images or posts on Facebook that evokes painful memories disappear into the sea of other ‘memories’ already uploaded unto Facebook; instead, there is an attempt to actively and automatically exclude certain memory objects from reappearing to protect people’s feelings. The issue of sculpting digital voids is that it is, therefore, not a neutral practice, value-free stagings of remembering; rather, they are ‘imagined’ constructions, and the ‘voids’ or spaces of invisibilities that these practices engender are moulded in and through algorithmically generated processes. These practices form part of an attempt to classify people’s past data on Facebook, their digital memory objects, a way of sorting their ‘memories’ based on the emotional response they are presumed to elicit. Thus, sculpting digital voids needs to be considered as entanglements of practices, imaginaries, management, classifications and power. As I argue later on, practices of algorithmically sculpting digital voids on Year in Review are problematic, especially as they fail to account for the complexities and nuances of remembering and its relationship to memory objects. Being encountered by past data, or what Facebook calls ‘memories’, elicits multiple responses. Sculpting digital voids may be an attempt to protect people from revisiting painful aspects of their past, but these practices also impose a framework on how people should respond to certain memory objects that are resurfaced on Facebook. Ultimately, sculpting digital voids does not account for and align with the richness and ambiguity of memory work. Before we explore the conceptual implications of sculpting digital voids, it is crucial to explore its functionalities in greater detail by looking at Facebook’s memory feature, Year in Review.

Facebook’s Year in Review and its ‘memories’

As already mentioned, Facebook’s memory application, Year in Review, underwent structural changes following a widely publicized complaint from Eric A. Meyer. The memory feature was introduced in 2012 as a means to revisiting the ‘best moments’ of the previous year. The feature would pop up during the final months of the year, and if one clicks on the feature, Facebook shows 20 highlights from one’s past year (Lang, 2012). It collated one’s archived Facebook pictures, videos and posts one has been tagged in, and general highlights, using an algorithm that categorizes said highlights based on the number of likes, comments and shares they got throughout the previous year. This means that posts, images or events that got a lot of likes are most likely to resurface. The app also collated and made a list of general topics that trended on Facebook over the past year (e.g. Facebook Newsroom, 2013). The app then generates a personalized montage video using users’ personal highlights from the past year, which one is encouraged to share with the rest of the Facebook community. These highlights, comprising posts, pictures or events, Facebook calls ‘memories’. As such, the memory feature on Facebook recirculates a user’s past data such as posts, uploaded images and events from the previous year and facilitates the re-engagement with these media objects as ‘memories’.

As a platform and digital archival structure, however, Facebook retains people’s (past) data indefinitely and permanently. As Year in Review, the platform’s memory feature works to resurface ‘memories’ back to users, reminding them of certain posts and images that they previously uploaded to Facebook, it is inevitable that the feature will occasionally resurface content that people would rather avoid. Such is the case with Eric Meyer. As a result, the app underwent structural changes to ensure that this would not happen again. It incorporated algorithms which sought to filter out ‘memories’ people most likely wanted to avoid (Dzieza, 2015). This process of filtering was explained in greater detail in the context of Facebook’s other memory feature, On This Day, which underwent similar changes. As Manohar Paluri and Omid Aziz (2016) explain, to optimize user’s experience of Facebook’s memory features, they developed automatic filters. These were meant to draw on metadata such as keywords within posts, images and comment sections to assess whether these ‘memories’ should be resurfaced or not. ‘Memories’ that included exes, dead relatives or family members listed as deceased would be filtered out and would consequently be resurfaced on Year in Review nor On This Day. Facebook also attended to, Paluri and Azziz claim, those ‘memories’, images or posts, that people had themselves blacklisted or did not wish to see resurfaced. This suggests, therefore, that the practice of sculpting digital voids is distributed among platforms and users.

It is crucial, however, to emphasize the active role the platform takes in rendering some ‘memories’ invisible. The effectiveness of Facebook’s Year in Review is predicated, at least in part, on the way they are able to channel a user’s ‘memories’ in the present. The feature makes visible people’s memory objects from the past year but actively seeks to render invisible those ‘memories’ deemed too emotionally offensive for a certain user. As I will discuss in further detail later, this has profound implications for subjectivity, how users experience the platform of Facebook and how they (re)engage with the past. Year in Review, this piece argues, attempts to actively ‘sculpt digital voids’ within people’s past data. That is, it seeks to algorithmically predict what a user wishes to see from the past year, navigates and manages what the user will see and ultimately has the potential to shape how the user experiences the digital past. But this algorithmic prediction also shapes what will become visible on Facebook’s memory feature and hence a user’s NewsFeed. Facebook’s practice of sculpting digital voids, therefore, has the capacity to shape a user’s experience of the past through rendering certain ‘memories’ invisible and, in turn, render other ‘memories’ resurfaceable. But what is meant by this notion of ‘sculpting digital voids’?

Sculpting digital voids and invisibilities

The concept of sculpting digital voids is drawn from Terry Eagleton’s (1976) Criticism and Ideology. In the book, Eagleton analyses the notion of ideology in relation to several authors, one of which is Joseph Conrad. He suggests that Conrad’s literary corpus is full of silences and silent figures which illuminate the ideological dissonances or tensions in his work (137). The interesting aspect of Conrad’s writings, Eagleton states, is not what he writes but what he does not write, what he chooses to leave out. This is particularly apparent in Conrad’s treatment of imperialism in Heart of Darkness, in which he neither explicitly advocates for the abolition of colonialism nor outright rejects imperialism but rather remains curiously silent. These silences, Eagleton argues, do not demarcate a space in which Conrad ‘should’ have written something but rather these silent spaces speak volumes in and of themselves. They are ‘determinate’ (137) not arbitrary or accidental. Eagleton argues that Conrad creates ‘a vacuum, sculpting a void’ (137) and that ‘it is precisely in these absent centres, which ‘hollow’ rather than scatter and fragment the organic forms of Conrad’s fiction, that the relations of that fiction to its ideological context is laid bare’ (138–139). In other words, it is through the sculpting of voids, and not in what is actually written, that Conrad’s ideological context becomes most visible. For Eagleton, therefore, it is a practice that actively seeks to actively render some spaces ‘hollow’ or ‘void’ in Conrad’s fiction.

This concept of sculpting voids reminds us that absences or silences are not merely residue from the fire of the already said and shown; they can be actively sculpted and shaped. For instance, Andrea Brighenti (2007: 328) states that the invisible is ‘what is here without being an object’. It is that which makes the visible possible, in the same way, visibility makes the invisible possible, and as such, they are dialectically interwoven. The visible makes the social intelligible to the human eye. Brighenti (2010: 39) even suggests that visibility is a kind of ‘social optics’, meaning what people consider to be the social is only that which is visible to them. For something to become visible and for us to see it, something else must become dark. ‘The invisible is’, in Brighenti’s (2010: 19) words, ‘intrinsic to the visible’. Similarly, Taina Bucher (2017) points out that invisibilities are fundamental to our experience of the world and other people. We may initially experience people through the way they dress, how they speak or the things they possess, but an essential factor in understanding people, Bucher states, is perceiving their unseen qualities: their moods, dreams, hopes and so on (32).

The role of invisibilities is equally prominent in a digital context. Daniel Neyland (2015: 127) suggests that knowing what data to exclude and delete is crucial for the development of algorithmic systems. Drawing on Hetherington and Lee’s work on the role of the blank figure, Neyland suggests that the practice of deletion ‘instantiates a kind of absent-presence’ (127), meaning that data that were deleted could not be completely obliterated but became ‘a zero which simultaneously signified nothing (data that no longer existed) and something (the ability of the system to delete and maybe even to prove the marketability of this system’ (127). Absences, for Neyland, signify both the absence of the data that have been deleted and the presence of the system that instantiated the absence. Sculpting digital voids represents a similar attempt at producing absence–presence, since it delineates the boundaries between what is considered resurfaceable ‘memories’ and not. This does not mean, however, that those images or posts that users blacklist in their preference settings are completely deleted off the platform (the question remains whatever is). Instead, sculpting digital voids seeks to avoid the resurfacing of certain ‘memories’ and thus navigate their particular absence–presence configuration. As such, the notion of sculpting digital voids does not suggest reified or stable ‘objects’ or ‘voids’ (this will be discussed in the penultimate section). It suggests a certain approach to people’s past data, how it should be approached, and what should and should not be resurfaced in the present. As such, the practice highlights how some past data are made invisible whilst also foregrounding the presence of Facebook’s politics of forgetting, that is, its algorithmic strategy which makes possible these invisible spaces. Just as Eagleton’s metaphor suggests a practice, so our notion of sculpting digital voids highlights a way in which some Facebook ‘memories’ are rendered unresurfaceable.

Of course, the importance of algorithms determining the visibility of posts and events shared on social media cannot be overstated (Bucher, 2012). However, the case of Eric Meyer and Year in Review also suggests the emerging yet problematic question of how invisibilities are actively configured on social media platforms. For John Thompson (2005), media such as television represented a new form of visibility. It constituted a subversive capacity to unveil formerly hidden practices such as the torturing of prisoners in US-run prisons in Iraq. ‘In this new form of mediated visibility’, Thompson argues, ‘the field of vision is no longer constrained by the spatial and temporal properties of the here and now but is shaped, instead, by the distinctive properties of communication media’ (35). Newer forms of visibility, such as the Facebook, NewsFeed or trending tweets on Twitter, which are algorithmically enabled, are similarly shaped by the properties of their media platforms. These forms of visibility, however, are also intrinsically entangled with the construction and management of invisibility or the sculpting of digital voids. The approach to people’s past data or their Facebook ‘memories’ on Year in Review demonstrates ways in which invisibilities are carefully sculpted and algorithmically managed, the making visible of certain images and posts, and the quiet concealment of those considered unresurfaceable. Sculpting digital voids, in this regard, is not the process by which people’s uploaded content sinks into the ever updating NewsFeed stream but rather the practice by which some ‘memories’ are rendered visible and some invisible.

The emergence of this practice, of course, inextricably linked with developments in retention and storage technologies. The archive in the digital age has, as Mike Featherstone (2000: 161) put it, ‘extended its wall’ to encompass not only significant historical events but most aspects of social life. In a sense, the archive was lifted out of its institutional, spatial and professional contexts, in which it had resided for decades and instead acquired the capacity to encapsulate all traces of our social being, including those mundane and everyday practices that often remained undocumented and unarchived. One of the differences, however, between the archive in the digital age and the ‘classic’ archive was the degree of permanence each one afforded. Whereas analogue objects such as documents, paintings, drawings, microfilm and audiotapes inevitably decayed over time, the digital archive afforded indefinite and seemingly infinite retention storage. For Andrew Hoskins (2013), the digital archive ushered in ‘the end of decay time’. He argues that the deterioration of objects afforded them value since they were perishable and therefore valuable (387). But with the rise of Big Data and post-scarcity culture, decay time was circumvented, and, instead, it became possible, and indeed desirable, for data companies to tap into the surfeit of people’s personal information (387–388).

But already at the turn of the millennia, Featherstone (2000: 170) suggested that digital archiving and permanent data retention would problematize the idea of the archivable:

If everything can potentially be of significance should not part of the archive fever be to record and document everything, as it could one day be useful? The problem then becomes, not what to put into the archive, but what one dare leave out.

These attempts at algorithmically excluding certain memory traces on Facebook’s Year in Review are, therefore, in one sense, necessitated effects of Big Data and post-scarcity. Unlike secrets, which arise in a situation where someone withholds information from someone else to control or dominate them, the act of sculpting digital voids emerges in situations when all information has become readily available and easily retrievable, where there is an excess of information and an overexposure to data. Even though practices of archiving are linked to relations of power, sculpting digital voids, on one level, suggests its own necessary nature. The documentation of all facets of social life inevitably results in some photographs resurfacing which evoke painful memories from the past. Such situations may, therefore, necessitate the sculpting of digital voids: blank spaces or invisibilities, in which this information can be contained and hidden from sight without being deleted – in the name of ‘protecting’ people’s feelings. These memory traces become figures of absent–presences (Neyland, 2015). Therefore, in addition to the threat of invisibility on Facebook (Bucher, 2012), there is an emergent threat of the ‘wrong kind’ of visibility on Facebook as well, one displaying painful and hurting memories.

As the Eric Meyer case indicates, the challenge for social media sites will increasingly become the management of people’s personal memories not just storing them. Sculpting digital voids is not merely to produce and manage (in)visibility but to actively carve out spaces of forgetting as well. In one sense, the exclusion of ‘memories’ containing flagged by Facebook’s filtering systems as containing dead relatives and ex-partners from the annual Year in Review montage video and On This Day reminders is needed and desirable for some Facebook users. Of course, the absence of such ‘memories’ cannot and should not be equated with straightforward forgetting. Yet by attempting to ‘hide’ certain memory traces from Year in Review, Facebook induces the forgetting of them by omission, downplaying their relevance and meaningfulness in the present in their annual review. As such, sculpted voids offer us a concept for analysing social media in a different way; in a way that examines the rendering of voids, invisibilities and blank spaces. This practice of sculpting digital voids can, therefore, be understood as part of an algorithmically enabled strategy of forgetting.

Memory/forgetting

As the Meyer case indicates, the end of decay time and the proliferation of digital data retention technologies have challenged our conceptualizations of memory and forgetting. Memory is no longer scarce, but it is readily and widely available through a vast array of social media sites. This permanence of digital retention has led some to suggest we are approaching, or have already reached, ‘the end of forgetting’ (Bossewitch and Sinnreich, 2012; Rosen, 2010). In his book, Delete: The Virtue of Forgetting in the Digital Age, Viktor Mayer-Schönberger (2009) states that in most of human history, the balance between remembering and forgetting has always tipped in favour of forgetting; most things were forgotten while it was special if something was remembered. With the advent of the digital age, however, this balance shifted and digital archives, databases and retention technologies have made it increasingly difficult for anything to be forgotten (2).

Even though Mayer-Schönberger’s notion of ‘the demise of forgetting’ in contemporary society is arguably overstated, it underscores the ethical and political implications of memory and forgetting. He states that even though the demise of forgetting has the potential to eradicate the painful and annoying consequences of forgetfulness (10), the problem is rather that

forgetting plays a central role in human decision-making. It lets us act in time, cognizant of, but not shackled by, past events. Through perfect memory we may lose a fundamental human capacity - to live and act firmly in the present. (p. 12)

Dodge and Kitchin (2007) similarly point to the ethical challenges that may arise from ubiquitous archiving and permanent data retention. They argue that these developments have had a disciplinary effect on people and that every action and event become digitally attached to individuals permanently (431). To counter these disciplinary effects, Dodge and Kitchin suggest that ethics of forgetting needs to be brought about, and ‘a range of algorithmic strategies could be envisioned’ to subvert the current trend and reintroduce forgetting into the equation (442). Some of these strategies include ‘erasing, blurring, aggregating, injecting noise, data perturbing, masking’ (442). These algorithmic strategies are meant to artificially induce forgetting by omission where natural forgetting has been circumvented and as such emancipate individuals from their pasts. Mayer-Schönberger (2009: 14) argues more specifically that the personal information we put up online should have an expiration date to better reflect the ‘finiteness’ of biological memory (14). In a sense, this would allow people to more easily move on from their past, and thus it would better facilitate self-reflexive reinvention.

The primary issue with notions such as ‘demise of forgetting’ and ‘the end of forgetting’, however, is they assume that forgetting is a uniform process. Such notions assume that the development of retention technologies has had the same effect on all areas where memory and forgetting are concerned, but the way people forget a grocery list naturally differs from EU’s ‘Right to Be Forgotten’ protocol or systematic repressions of memory often seen under totalitarian regimes. As Connerton (2008) points out, ‘forgetting’ is an umbrella term comprised of multiple types or acts of forgetting, which overlap and interconnect. Connerton’s fifth type of forgetting, ‘forgetting as annulment’, ‘flows from a surfeit of information’ (64). He predicts that ‘the concept of discarding may come to occupy as central a role in the 21st century as the concept of production did in the 19th century’ (65). This emerging centrality of discarding as a form of ‘forced’ forgetting can be seen on Facebook’s two memory apps, Year in Review and On This Day. Sculpting digital voids echoes this need for the annulment and discarding or at least the invisibility of some ‘memories’. Rather than witnessing the end of forgetting, we instead witness the production, dissemination, selection, management and realization of forgetting on social media – in other words, forgetting in its sculptedness.

From the outset, then, sculpting digital voids on Year in Review may seem like one of Dodge and Kitchen’s (2007: 431) ‘algorithmic strategies’ that resist and subvert the demise of forgetting and become a way in which to rebalance the relation between remembering and forgetting – in short, a practical strategy to emancipate people from their painful pasts. However, artificially inducing forgetting is as complex and messy a process as biological forgetting, albeit different. Whereas biological forgetting happens naturally and inevitably, strategies of artificial forgetting need to be envisioned by someone; they need to be created in a particular way. Forgetting as annulment, as opposed to naturally inevitable lapses in memory, is an ‘act’ of forgetting, a coping strategy to navigate the surfeit of information in society and a service provided by social media sites to help users navigate the surfeit of their own personal memories. Algorithmic strategies that seek to induce forgetting by annulment or discarding are undergirded by specific assumptions about the world.

Even if algorithms on ‘Year in Review’ do prevent people from reliving the painful memory of a deceased loved one or messy break up, it is crucial to ask the question: ‘who decides?’ As memory is never divorced from a remembering agent, Paul Ricoeur (2004: 3) asked the simple question: ‘Whose memory is it?’ When one asks ‘who’, it becomes evident that the question is interlaced with issues of power and authority. As we mentioned earlier, Facebook actively participates in the sculpting of digital voids on Year in Review. We, therefore, need to ask: who forgets? Who chooses what is to be forgotten? The algorithm or the remembering agent? When should it be forgotten? How should it be forgotten? Facebook’s Year in Review seeks to filter and excludes certain ‘memories’ based on metadata such as keywords and thus preventing situations like with Eric Meyer. As David Beer (2009: 996) argues, ‘The power of the algorithm in this instance is to shape auditory and cultural experiences’. The question of what ‘memories’ remain invisible and which ones are resurfaced is inextricably interwoven with the power of the platform and its algorithms ‘to make choices, to classify, to sort, to order and to rank. That is, to decide what matters and to decide what should be most visible’ (2017: 7). The platform, therefore, has the potential to shape what is remembered and considered meaningful in the present as well downplaying other ‘memories’ that are considered too painful to be resurfaced.

In his book, Mayer-Schönberger (2009) suggests a practical way in which forgetting can be reintroduced into the equation, but he fails to address the issue of power that arises as a consequence of that reintroduction. It is not merely enough to address how we can resist the demise of forgetting and algorithmically reprogram forgetting into our online applications; we also need to address the question of ‘who’ is to facilitate this process and how it will be implemented. Featherstone argued that the problem with ubiquitous archiving was not what one included but what one chose to leave out. However, the problem with algorithmic strategies such as those mentioned by Dodge and Kitchin (2007: 442) or Mayer-Schönberger (2009: 14) is twofold: ‘who’ chooses to leave ‘what’ out? Social media sites such as Facebook are in a position of power to shape people’s experience of their digital past, and, therefore, the question must also be extended to the way sculpting digital voids occurs on Facebook. This practice, the production of invisible spaces, is never a neutral act. Even if it subverts the demise of forgetting, sculpting digital voids is embedded in a particular mode of understanding and sorting people’s (past) data on Facebook; indeed, it is a way that classifies past data according to their resurfaceability or unresurfaceability. It is, therefore, imperative that sculpting digital voids be considered in relation to power and as constituting a mode of classification.

Sculpting digital voids as mode of classification

Year in Review do an annual list of trending topics and collate this into a top 10 list. In 2016, the US presidential election unsurprisingly topped the list. The 2016 edition of Year in Review also included for the first time a top-10 list of global live videos on Facebook (Facebook Newsroom, 2016). In the 2013 edition, the Facebook app created a list of ‘Top Life Events’, consisting of the events people most frequently added to their Timeline. The inputs for the list were mined from the places and events that generated most posts on people’s Facebook profiles and walls, which Facebook then put into a national and global ranking system (Facebook Newsroom, 2013). On their list of top-posted and buzz-generating events, ‘Added a relationship, got engaged or got married’ was number 1; ‘Traveled’ and ‘Moved’ were numbers 2 and 3, respectively, and interestingly, number 4 was ‘Ended a relationship’, and ‘Lost a loved one’ was number 8 (Facebook Newsroom, 2013). Even though the exact order of the events inevitably fluctuates from 1 year to another, the list is a general indication of the kind of events and experiences that are central to people’s everyday lives.

The reason I refer back to Facebook’s ‘Top Life Events’ is to twofold: firstly to emphasize that sculpting digital voids does not produce reified objects or ‘voids’ and, secondly, showcase that Facebook’s modes of classifications are inextricably linked with process of constructing and managing invisibilities. Sculpting digital voids is a form of algorithmic classification, a way of defining and, therefore, rendering some ‘memories’ unresurfaceable as others are considered ‘top life events’. Sculpting digital voids is a way to categorize and sort people’s past data, their Facebook ‘memories’ – that is, excluding break ups and relatives listed as deceased from the annual Year in Review – and as such, they do not constitute a neutral domain. As Geoffrey Bowker and Star (1999) remind us, classification is fundamental to human perception and the sense-making processes of reality. Bowker and Star (1999: 10) define classification as the ‘a spatial, temporal or spatio-temporal segmentation of the world’. David Beer (2013: 43) even calls it ‘a kind of unconscious ordering force’. Processes of sorting and classification do not only occur on a political and structural level, but are deeply enmeshed in the everyday, to such an extent that they may become unconscious, invisible and taken for granted. The human inclination to classify is embedded in all facets of life.

Yet, even though classification is a deeply embedded process, it does not always remain a neutral one. The categories that enable and are in turn enabled by classification are constructed by someone. Even though algorithms can be thought of as neutral mathematical procedures, transforming massive data sets into intelligible outputs, they are always created by someone (Mittelstadt et al., 2016). Their classificatory power is often underpinned by a certain way of looking at the world, and these invisible choices are often also embodied in visible practices, processes or material objects. For Safiya Noble (2018), the algorithms underlying the Google search do not mainly showcase their ‘inadvertent cruelty’ but rather the hidden biases in those that developed them. More broadly, David Beer (2013) suggests that the circulations of data and algorithmic processes, which shape popular culture and people’s cultural tastes, reflect someone’s vision of the world. He proposes the concept of ‘classificatory imagination’ to emphasize and encapsulate the various, entangling processes of classification, both human and nonhuman, that are at play on social media platforms (46). Given the entangled process of sorting, classifying and hierarchization, piercing the materiality of the algorithm to see its underlying logic(s) is severely problematic (Burrell, 2016; Pasquale, 2015). As Mittelstadt et al. (2016: 6–7) point out, the construction of algorithms often involves teams of developers that do not possess a holistic understanding of the process they are involved in. Even though algorithmic processes are obscure, complex and seemingly impenetrable, Beer’s concept reminds us that modes of classification do not just happen – they are ‘imagined’ by someone, undergirded by certain assumptions about the social world.

To what extent, then, is sculpting digital voids a form of classification? Regardless of intention (e.g. protecting people’s feelings), attempts to algorithmically regulate what data may resurface on people’s Facebook Year in Review are rooted in a specific way to relate to and classify a user’s ‘memories’ on Facebook. As Paluri and Aziz (2016), Facebook’s Year in Review and On This Day both utilize filters which pick up on keywords in users’ posts, images and comment sections that suggest potential ‘memories’ containing exes or dead relatives. Implicit in this is a notion of ‘good’ and ‘bad’ memories, past data which can be resurfaced and past data which should not be. For Eric Meyer, for example, being faced with images of his recently deceased daughter was very painful, an experience he would rather have avoided. Similarly, others would rather never see their ex-partners again, or any of the Facebook memories attached to them, in which case these algorithmically sculpted voids are beneficial. As Facebook’s ‘Top Life Events’ list suggests, these occurrences are unfortunate yet natural and inevitable parts of everyday life and it is, therefore, questionable whether they should always remain invisible or one should be induced to forget them by omission.

Yet, by seeking to sculpt voids around dead relatives and ex-partners, Facebook assumes all users will react to these memories in a broadly similar way that Eric Meyer did. As remembering is inescapably rooted in various emotional, socio-economic and historical contexts, the act of sculpting digital voids constitutes a simplification of the various emotional responses memory objects can evoke. For some, a dead relative may not induce numbing pain but a deep appreciation of moments shared or a life shared; for others, an ex-partner may not be the devil but rather a friend that remains close even though life took them in another direction. And drawing on Facebook’s ‘Top Life Events’, number 2 (‘travelled’) and number 3 (‘moved’), for example, are not always positive as numbers 4 and 8 are assumed to be inherently painful. One may have been on a vacation which turned out to be catastrophic or one had to leave one’s dream home. Nonetheless, ‘Year in Review’ has decided that these memories should still be excluded regardless if some people feel otherwise.

This issue, however, of simplification, is a discriminatory quality of algorithmic categorization in general. Mike Ananny (2016: 103) states that ‘algorithmic categories raise ethical concerns to the extent that they signal certainty, discourage alternative explorations, and create coherence among disparate objects’. Algorithmic processes infer meaningful categories through cementing similarities and limiting the field of possible alternatives. Although algorithms are effective at recognizing patterns and similarities in data, the way things are categorized can lead them to become ‘narrowly construed’ (Ananny, 2016: 103) and understood through simplified models. Algorithmic categories also signal certainty in a way that essentializes events, construing the resurfacing of dead relatives as an inherently painful encounter, regardless of the rich context through which such memories must necessarily be interpreted. As such, algorithmic classifications face the challenges of simplification and essentializing challenges that Facebook’s memory apps also face.

Through these acts of sculpting digital voids, a classificatory imagination emerges: Facebook’s Year in Review has imposed an algorithmically enabled, normative and objective framework onto their users in how to relate to certain Facebook ‘memories’. There is an underlying logic that algorithmic categorization is a compatible facilitator of people’s digital memory objects. It is evident that the way ‘memories’ are sorted and classified on Facebook’s memory apps is not a neutral nor objective process. At one level, one may even suggest that sculpting digital voids forms part of what William Davis (2015) called ‘the happiness industry’, power structures, which shape behaviours and notions of happiness in relation to people’s data. But I want to argue that sculpting digital voids also demonstrates that Facebook’s system of classification does not account for the dynamism of lived experiences nor do they sufficiently recognize the complexities and ambiguities of remembering the past, the various feelings memory objects may induce and how people may ultimately react to them in the future. These practices are indicative of a classificatory imaginary, an attempt to better manage people’s data and digital memory objects. They are an ‘imagined’ politics of classification on Facebook; a politics undergirded with invisible assumptions about memory and forgetting. Thus, the concept of sculpting digital voids offers us ways to examine the way visibilities and invisibilities are managed and navigated on social media and to analyse the role of classification in digital contexts. Seen through this conceptual lens, classification is not merely about sorting out objects (or memories) in relation to pre-existing categorizes but to determine the state of visibility or invisibility these objects will obtain in a future present.

Conclusion

With the ever-growing presence of algorithms in society, the need for further analysis of their effect on memory and forgetting grows accordingly. The rise of digital archiving has brought about unprecedented opportunities and challenges for individuals and whole societies. Developments in retention technologies, such as digital memory apps, impel us to critically examine their nature and function in relation to how we remember and forget. The number, availability and general use of memory apps in society are on the rise. One can mention Timehop, Facebook’s Memories, Memoir, Snapchat Memories and Google Photos. Timehop, for example, had 7 million users check their app daily already in 2015 (Kosoff, 2015), and as of 2018, Timehop also has over 500,000 followers on Facebook. Most of the big social media platforms have incorporated features into their functionality that mediate and capitalize on people’s personal memories; even smartphone software (e.g. the iPhone) have embedded processes that algorithmically curate and mediate people’s photos as memories, automatically reminding people of photos they have taken x amount of time ago. That a market for digital memory objects has emerged is unsurprising, given the substantial amount of personal data people store online. The rise of memory apps and the increasing use of them suggest that digital memories are not merely a hype but a broader digital and societal trend.

My intention has not been to treat Facebook’s Year in Review as a closed-off, isolated or self-contained digital archive. Digital archives are often collections of multiple types of content, accumulated from multiple various sources; products of ‘accumulated constituent profiles organized through metadata’ (Beer and Burrows, 2013). Rather, understanding ‘Year in Review’ as an archival structure, I used the Facebook memory app to frame an examination of the conceptual implications of digital archiving on memory, forgetting and ultimately the self. But to increase the potential for a more nuanced understanding of memory apps as a field of critical study, future research could examine an array of memory apps to better understand their contextual embeddedness in various assemblages and their relation to memory and forgetting.

The notion of sculpting digital voids, which I argue is central to Facebook’s politics of forgetting, can become a valuable, conceptual tool because there is need for additional study into the roles played by invisibility and forgetting on social media and the role played by algorithms in sorting people’s past data. Sculpting digital voids has become important in a context in which there is an overexposure to personal data and memories. And since the amount of data archived on various social media is bound to grow over time, it is conceivable that platforms will take on an increasingly active role in the remembering of the past. The practice of sculpting digital voids is ‘imagined’ by someone. They are a form of classification and must, therefore, be examined in relation to power. Through these practices, the ideological context of Facebook’s ‘Year in Review’ emerges – a politics of forgetting. The concept is a reminder that invisibilities are carefully crafted and are undergirded by a ‘classificatory imagination’. As digital archiving becomes increasingly embedded in everyday life, the need to better understand these practices of sculpting digital voids and its effects on memory, forgetting and the self becomes increasingly immediate.