Coding the Dead: An Interview with Tamara Kneese
Tamara Kneese | Death Glitch: How Techno-Solutionism Fails Us in This Life and Beyond | Yale University Press | August 2023 | 272 pages
When the living log on and search for the dead, what shows up? Search for a deceased artist, scientist, or politician of some notoriety and your screen will populate with belabored obituaries, Wikipedia entries, and web forums reflecting the nearly monastic dedication of enthusiasts scattered in physical space but connected online, sharing in the sacred, secret rites of the phenomenon called internet fandom. The dead may loom large in cyberspace, but the length of their digital afterlives is entirely uncertain.
In Death Glitch: How Techno-Solutionism Fails Us in This Life and Beyond, technology scholar Tamara Kneese, director of Data & Society’s Algorithmic Impact Methods Lab and former green software researcher at Intel, explores the precarity of our data and digital selves. Kneese seeks to show how individual death represents a “glitch” in the platform economy and logic of venture capital. Call it a root error or forever bug: despite the profusion of technosolutionist and transhumanist schemes to escape death, an individual user’s demise reflects an essential precarity in platforms themselves.
What is to be done with the data of our deceased is a matter of policy for platforms and practice for loved ones. The social media profiles of the dead, for example, represented a problem of governance for companies like MySpace and Facebook, but they are equally important objects in the grammar of grieving and remembrance. The fragility of human life, Kneese suggests, is mirrored in our digital remains. In the end, digital death ripples through communities of kin and care, who shape what becomes a legacy.
For the famous and un-fêted alike, what happens to our data when we die is a pressing question, and Kneese presents readers with possibilities in episodes on social media memorials, illness blogs, digital estate planning, and the haunting quality of a deceased’s smart devices. The overall effect of Death Glitch is to puncture any inflated sense of historical difference prevalent in the digital age. The terabytes of data we produce every day, the tools we use to publicize the pulse of our lives, do not constitute fundamental breaks with the human condition, but reveal an essential uneasiness in confronting it.
Social media is no panacea for senescence. A far cry from a comfortable sinecure in historical time, digital life — the bricolage of profiles, likes, favorites, notes, and memes we leave behind — can and may well disappear with a wrong swipe, a server crash, or the banal filings of corporate bankruptcy. Glitches disaggregate digital selves into strings of binary in the bitstream. Who knows where we go? Perhaps our datastreams will feed large language models, as corpses once fed soil. Is a modern digital life merely that of a microserf, indistinguishable that of the long-dead peon lost to records of the human past? In the sweep of historical time, this may be nothing but a narcissism of small differences. Here and now, however, Kneese shows how death represents an intractable paradox in the platform economy.
This conversation has been edited for brevity and clarity.
Jacob Bruggeman: When humans die, what happens to all the profiles and data that made up their digital identities?
Tamara Kneese: It depends! And that’s one of the reasons why communicative traces, or what I refer to as the slippage between real time personal data that serve a utilitarian purpose and the aggregated digital remnants of a life that become digital remains, create so much anxiety for individuals and their surviving loved ones. In some cases, social media profiles, personal websites, and email accounts continue on as before, perhaps accruing spam messages or attracting hackers while gathering pixelated dust. Other times, they can be placed in a memorialized state and marked as sacred spaces for remembrance.
This relies on platforms having a formalized memorialization policy and the ability to operationalize it. But it’s also possible that the platforms and websites that digital remains are attached to will disappear altogether, much in the way that journalists and other writers lose the traces of their creative work when magazines or news outlets go belly-up. Large platforms like Facebook were among the first to have memorialization policies, but even those measures are imperfect because they often rely on individuals to plan for their own deaths and appoint legacy stewards ahead of time.
A number of startups around digital estate planning emerged in the mid-2000s as a way to help fill in the gaps left by Trust & Safety policies and the legal system for passing down digital belongings, but as I found in my research, the majority of those startups died long before they were able to facilitate an intergenerational data transfer. With generative AI, the idea that one can be revived through data has become mainstreamed. So with these systems, it is possible that the data of the dead can be used by the living and reconfigured in new ways.
JB: Your book argues that social media platforms fail to grapple with the processes of grieving, memorialization, and sacred ritual that define how most of us cope with the death of those we loved. One might argue this is simply a problem of user experience design. In that sense, it can be corrected, as with Facebook. But your book points beyond the architecture of platforms to their animating impulse: profit. Can you explain?
TK: Death is an event that yields interaction and engagement, which is a good thing from the perspective of social media platforms. To some extent, the capacity to memorialize and collectively mourn the dead may keep users interested in an aging platform that has fallen to what Cory Doctorow calls “enshittification,” or when the platform’s attempts at maximizing profit undermine the best interests of its users. It might be harder to leave the AI generated content on Facebook or the Nazi shit on X if you’re sticking around to visit the profiles or tweets of dead loved ones.
Despite the fact that death should be an obvious use-case at this point, many companies fail to remember the profiles of the dead when they decide to deactivate inactive profiles or free up inactive user handles. So while death might to some extent offer platforms a way of keeping reluctant users in the network, death can lead to other issues, like the cost of data storage and the threat of security breaches or hacks. For platforms, mourning and memorialization may offer a gateway into surveillance or engagement, but it doesn’t necessarily translate into the bottom line.
JB: The result of that laser-focus on profitability is the book’s central concept, the death glitch. Surveying social media memorials, illness blogs, and digital estate planners, you chart so many of the ways grieving has gone online. What does the death glitch explain about our shared meanderings through the digital graveyard?
TK: Since the earliest days of the internet, people have found ways to mourn the dead online. This was true on early mailing lists like Future Culture and early virtual communities like The WELL (Whole Earth ‘Lectronic Link). But with the rise of social media and the value of social data for advertisers and government agencies alike, websites like Facebook became spaces for hanging out in everyday life and, in moments of individual death, places for memorialization or communication with the dead. But because memorialization is often mediated by commercial platforms, and by startups that tend to disappear almost as quickly as they appear, people cannot depend on the longevity of digital remains.
No matter how many solutions to mortality emerge, devices and websites die in the same way that people do. And it is the moments of breakdown in the face of death that provide a glimpse of the networked, relational labor behind digital production. There is a fantasy now, with generative AI in particular, that we can safely become post-user, that capitalism can perpetuate itself through bots alone. But there are all of these points at which it becomes clear that humans are always there behind the scenes. So many cases of faux automation are in fact simple stories of outsourcing.
JB: There are a number of ways scholars and critics have styled the moment when tech breaks down. What’s the unique payoff of the “glitch” as an analytical category?
TK: The glitch is not just a point of breakdown, it is also an unexpected and potentially transformative moment, a kind of aesthetic surprise. Legacy Russell frames glitching as action, as a movement. The failure to perform as expected can be strategic and productive in its own right. This is why at the end of the book, I refer to gig worker-led movements around death care demands as glitching the system, a way of using GoFundMe as a way of calling attention to the need for mass social movements and systemic change.
I’m also drawing on feminist, trans, and queer media theory accounts of the glitch as a way of extending liberal human subjectivity. The STS scholar Anne Pasek argues that glitch aesthetics reveal non-human agency in artistic production. Whit Pow uses glitches as a way of showing presence through absence when it comes to trans histories of computing. How do you capture the history of moments and entities that haven’t been fully saved by the archive? That is a methodological question that haunts the book, particularly as screenshots became my imperfect way of archiving fleeting field sites from the ephemeral web.
JB: The story of memorials to the dead on social media platforms is largely a story of organizing among grassroots users to demand new policies to protect and serve posthumous profiles. When did user-led movements to establish new digital death practices emerge?
TK: Facebook was created by elite college students for their own kind, and so death was not baked into its design. It took the mass shooting event at Virginia Tech in 2007 to change Facebook’s relationship to death, and to make death visible on the platform and to users. At the time, Facebook’s standard policy was to deactivate the profiles of dead users if family members supplied an obituary and death certificate. Otherwise, the profiles stayed online as before. But because of the major media event of the Virginia Tech shootings and journalists’ use of Facebook to glean information about the victims, Facebook said they would deactivate the profiles of all the victims after a 30-day period, which was their standard operating procedure at the time. To the company’s surprise, friends and family members petitioned Facebook to maintain the profiles of the dead as virtual shrines. It was the first time that death became an obvious use case, and Facebook had to learn how to grapple with that unexpected reality.
JB: How are those movements faring on platforms today?
TK: People are still fighting with corporate bureaucracies after their loved ones die. There are so many platforms and each has millions if not billions of users all over the world. Local laws and cultures mean that a one size fits all approach to memorialization doesn’t work. Users might push back against policies they find restrictive. For instance, should a platform ban deepfake versions of dead people outright, or should they allow grieving parents to create deepfake version of their dead children? It’s macabre, but this is the sort of issue that platforms, content moderators, and users, including the bereaved, face.
There is also the notion, with generative AI in particular but really with the data of the dead in general, that there is a potential to extract patterns of information and prediction that will automate other forms of labor. For example, the fantasy that we can use deepfake versions of actors or generative AI to create new books by dead authors is at its heart a fantasy about cutting labor costs. And the desire to harness the data of the dead for commercial, labor-replacing purposes has been met with resistance from unions like SAG-AFTRA.
JB: Some platforms and technologists are trying to leverage mountains of data to resurrect the semblance of a person from the digital graveyard. Training AI on the texts and emails of the deceased can reproduce some aspects of speech and affect that rhyme with our memory of them. Proponents of this technology argue it can help with the grieving process. But no technology is without costs.
TK: The idea that one can approximate a human subject, or a soul, through data holds some fascination. For some, it might offer comfort in the face of one’s own mortality or the loss of a loved one. There are many problems with such facsimiles of the dead, however, including their relationship to proprietary platforms, their underlying infrastructures and maintenance requirements, and ethical and legal questions about consent and authority. Who should have the right to use the data of a dead person to revive a chatbot version of them? A close family member, a friend, the executor of an estate, or perhaps an employer?
That person’s data is an entire constellation of interactions and relations with others; personal data is still visible and meaningful through networked communication. Would an individual feel comfortable having their data used in this way, or would their loved ones? There is also the problem of grieving a second time, as many of the systems that produce such simulations of the dead may depend on short-lived companies or technologies that soon become obsolete.
Finally, data storage requires the energy, water, land, and other resources needed to sustain massive data centers. The financial cost and environmental toll of such data ghosts are real considerations.
JB: You reference the concept of maintenance here, which is having something of a renaissance.
TK: It is somewhat amusing to me that Stewart Brand’s new book is on maintenance, taking a kind of DIY, tinkering approach to the subject. There has been a long-standing infrastructural turn in STS and media studies, where we acknowledge that the internet is made of stuff, from undersea cables to the human workers who make large platforms function.
Shannon Mattern’s essay on maintenance and care remains one of my favorites because it cautions against romanticizing repair. For whom is maintenance and the physicality of infrastructure a romantic adventure, and for whom is it a necessity and a drudgery? That is something Death Glitch attempts to bring to the fore: what are the various layers of maintenance and care required to preserve digital remains, and what kinds of political and economic choices are bound up in this kind of work?
JB: Data extraction from historical records and the residue of countless actions online might mean that we are all connected, in some small way, to almost any large-language model. On one hand, as much of the pushback against Chat GPT has illustrated, there’s a major privacy concern here. On the other, there’s something profoundly collective and empowering about knowing pieces of ourselves will reappear in the digital world long after our demise. How should we negotiate that tension?
TK: This is an intriguing question and I think it gets at the heart of why I wrote Death Glitch. As I said in a recent essay for The Baffler, I’m somewhat sympathetic to Ray Kurzweil’s quest to revive his long dead father through his physical, paper-based records that he keeps in a storage unit, to build a chatbot version of him even if it’s dependent on proprietary software that the Kurzweil family ultimately does not control.
ChatGPT is certainly people, Soylent Green style, and it unfortunately is being used as an excuse to cut human labor costs. The idea that someone would be monetizing collective human intelligence and benefiting from it while also immiserating various workers, from annotators to writers to teachers, is quite grim. The dead do their own work and live on in many ways, and digital remains are in many respects an extension of other forms of mediated legacies. It’s one thing to leave behind snippets of text that make their way into the ether, which might be comforting for some, but it’s another for that text and your entire corpus to serve as the basis for extraction. After all, capital is dead labor.