Grieving Machines

The pitch is simple: drafting an obituary for a loved one can be a difficult and emotional task, so why not outsource the work to a computer? Recent advances in generative AI have made it a cinch. Just plug in some basic biographical information, a few major life events, a hobby or two, and an advanced language model, having already scraped the internet to construct multidimensional analyses of linguistic patterns, will pop out a concise and respectable obit.

Such is the promise of “Finding Words,” a new tool from Empathy—a tech start-up that helps the bereaved with, among other grim tasks, funeral planning, “processing grief,” and, yes, drafting obituaries. It takes on average 540 hours to settle a deceased’s affairs, Empathy reports, so imagine if AI could save a few here and there?

Auto-obituaries generated by the likes of Empathy aren’t going to find their way into the New York Times or the Washington Post any time soon. They’re far more likely to wind up in local newspapers, the fees they garner helping to offset the costs of declining ad and subscription revenue as mourners pay. Or perhaps they’ll end up online, on websites like Legacy.com,which hosts death notices and obituaries for just $99 and offers to help place obituaries for additional fees at close to three thousand partnering newspapers. (Legacy, coincidentally, also offers an auto-obit service.) In any case, the obituaries Empathy automates stick to the same prosaic and formulaic structure that journalists of the death beat have settled on: [INSERT NAME], [INSERT THE DECEASED’S TOP ACCOMPLISHMENTS], died on [INSERT DAY] at [INSERT LOCATION OF DEATH]. They were [INSERT AGE].

The task is easy for tools like Finding Words and a cakewalk for new, larger models like OpenAI’s recently released GPT-4. The program may indeed help grieving people manage the stress and workload presented by a recent loss. It is true that the labor imposed by loss and bereavement places unwelcome burdens on working people: as Empathy points out, “47 percent of employees indicate their performance was negatively affected by loss.” Automated writing may be an appealing option, too, for journalists weary of reporting on death on deadline. There is no denying that Large Language Models (LLMs) and generative AI are powerful, but powerful technology does not unequivocally demand its use.

Obituaries, in distilling a lifespan to a few hundred words, mirror contemporary values; they write and reinforce scripts for what constitutes a “good life.” Language models trained to recognize and parrot the linguistic patterns of old obituaries are more likely to reproduce those scripts, inadvertently projecting forward the values of the past. While technology may enable broader representation of everyday people in obituary pages, the specific technology of LLMs risks reproducing the biases that have emerged in obituaries—like the veneration work and the single-minded attention to “notable” achievements. Before we embrace the age of auto-obituaries, we must think more deeply about the submerged political content of the form itself.

Contemporary obituaries in the United States bear limited resemblance to their ancestors. Benjamin Franklin’s death in 1790 prompted the Pennsylvania Gazette to publish no more than a cursory death notice and short doctor’s note grieving its editor. Franklin’s obituary in another paper, the Federal Gazette, ran a lengthy sixty-nine words. He was, the paper noted, a “FRIEND OF MANKIND” whose “singular abilities and virtues” had long been proven to the world. Franklin was survived by two children, but this did not seem to warrant attention. Neither did his invention of the lightning rod, bifocal glasses, or his co-authorship of our founding documents. It was, in the editors’ view, “impossible for a newspaper to increase his fame” and therefore not worth a longer treatment.

Through the eighteenth century, papers continued publishing relatively undetailed death notices. The work of memorializing the dead lay elsewhere, with the eulogist, who, at the funeral, would ideally go long on the deceased’s wealth and Protestant devotion. Obituary pages had not yet joined those two tasks. Generally, notices were only given if prominent community members died; in the case of Franklin, his good deeds and noteworthy contributions were presumed to be common knowledge. There has been limited study of these early American obituaries, but it comes as no surprise that they overwhelmingly focused on the white, male, land-owning elite.

By the Civil War, American obituary writers began to find more words. Mass death and what the historian Drew Faust termed “shared suffering” left Americans scrambling to remember those lost on the battlefield, and newspaper death notices of this time were occasionally accompanied by facts of the life, family, and career of the deceased. As the historian Janice Hume points out in her book Obituaries in American Culture, some of these more detailed Civil War obituaries and death notices were submitted by loved ones of the deceased for publication, perhaps through a process not wholly unlike the way Empathy users might submit auto-obituaries to their local papers. Others were penned by journalists.

In the years after the Civil War, Hume notes, obituaries began quoting less and less from the Bible, opting for celebration of the deceased’s national service and commercial successes. Major J. L. Perkins, for example, an officer during Sherman’s March to the Sea, was remembered in Iowa’s the Hawk Eye following his death in 1871 as “a brave and faithful soldier, esteemed and beloved by his comrades,” a man who “carried the same qualities into private life, and has a huge position among the business men of the city.” Perkins’s service came first, his private life second, and his surviving family third. Women, insofar as Civil War-era obituaries deigned to remember them at all, were nearly always closely associated with their husbands and domestic work.

Technology, too, played a role. Part of the reason newspapers could begin to publish longer, more detailed obituaries was the wide deployment of the “penny press,” which had been invented and first used by a New York paper called The Sun in 1833. It vastly reduced printing costs, enabling papers to print and distribute at ever increasing volumes. Common schooling, increasing literacy rates, and expanding railroads and telegraph services all contributed to wider readership: between 1840 and 1850, the U.S. Census reported the number of newspapers in the United States increased from just over 1,631 to 2,526.

As America industrialized and professionalized into the twentieth century, obituaries increasingly came to privilege workplace achievements, as the sociologist Bridget Fowler argues in her book The Obituary as Collective Memory. In 1900, Fowler says, aristocrats were the most common subjects of New York Times obituary, “evok[ing] a world where destinies are fixed by birth.” But that began to change by the 1940s: as the middle class began to expand, high-achieving professionals and managers began earning a place in the New York Times obituary pages; these remembrances highlighted the careers and industriousness of their subjects. But, as Fowler writes, “no obituaries appeared of working-class men and women.” These trends in the New York Times set standards and models for other obituary writers, journalists or not. Loved ones paying to publish shorter obituaries had to choose which details of their loved ones’ long lives were worth publicizing. Work, wealth, and contributions to American industry seemed natural choices.

Today, the form is thoroughly standardized. The Associated Press offers a lengthy list of guidelines. For example: use “death” and “die,” never euphemisms like “passed on.” Implying the existence of an afterlife, it seems, is too spiritual for the allegedly objectivist practice of obituary-drafting. Sometimes, when obituary writers do stray from accepted formula, controversy ensues.

Consider in 2013, when the New York Times ran an obituary for Yvonne Brill, a pioneering female rocket scientist. They chose to eschew the highly coded and standardized first line of obituaries. Instead of the name, age, and date of the scientist’s death, the Times quoted her son saying that “she made a mean beef stroganoff.” It seemed fairly obvious to readers at the time that this choice—to center a woman’s domestic labor over her pathbreaking research—was sexist, emblematic of the prejudices that have and continue to make the hard sciences overwhelmingly male fields hostile to the contributions of female researchers. The event also reminded us that obituaries are not politically neutral documents.

That memory work is inherently political is hardly a novel suggestion. Tragedy invites reflection; reflection invites contestation. Well-studied are the processes and conflicts over the accuracy, distortion, or erasure of memory in post-Nazi Germany, in towns of the American South littered with Confederate monuments, in Argentina after the military junta. Often, such studies center on public memorials, reconciliation policies, or prominent art installations and projects.

Among academic historians and sociologists the study of the ways groups remember their past has blossomed into a field some call “memoriology.” Memoriology entered history and sociology departments in the early 1920s, after Maurice Halbwachs published “On Collective Memory.” Halbwachs’s contribution—the concept of a kind of memory of past trauma and historical inheritances that belonged not just to individuals but was shared across groups—soon became the subject of heated academic debates. Responding to Halbwachs, some have cast doubt on the possibility that memory—an undoubtedly individual and personal mental state—could ever truly be “collective.” Others, seeking to defend Halbwachs’s formulation, have set out to identify the ways groups of individuals might use memorials, museums, symbols, or the practice of popular history to produce shared and selective conceptions of the past.

Fowler is among few sociologists to apply Halbwachs’s coinage and the discoveries of memoriology to the death pages of newspapers. One of the challenges of defining collective memory is locating what appears to be an individual mental state—memory—in groups. In the case of mass death, we erect monuments and host vigils. In the case of individual death, we draft obituaries. Fowler argues they serve as a crucial technology bridging the gap between individual loss and community memory. “Obituaries,” she writes, “are important because they seize a person from singularity and canonize or consecrate them.” One need not look further back than 2020, when the New York Times published a list of the first one hundred thousand people to die from Covid-19 in the United States, memorializing the pandemic as a moment of “incalculable loss.” The feature sought to locate individual names in a world-historical event, to unify the loss of many in the public record.

Whatever their intentions, when individuals pay to publish their loved one’s obituary, they code their grief as public loss. Those fortunate enough to be identified as noteworthy by major newspapers might receive long, detailed treatments of their lives, sent off to millions of readers. When journalists ask and answer what from any individual’s past makes them worth five hundred words in the New York Times, they ask and answer inherently ethical and political questions. By memorializing women as beef stroganoff-makers, working people for their contributions to industry, and service members for their heroism on the battlefield, contemporary obituaries render some lives memorable and indeed specify exactly which details of a long and varied life are worth remembering. Even the most transgressive obituaries—those that record hiccups and moral turpitude—still rely on productivity and achievement as metrics of value.

Of course, not everyone is quick to acknowledge the prejudices of the form. Bill McDonald, the editor of the Times obituary desk, attributes the overwhelmingly white, male annals of their obituary pages to historical inequity—which journalists were helpless to change. “The people whose obits are appearing in our pages now largely shaped the world of the 1940’s, 1950’s and 1960’s, and the movers and shakers in those eras were predominantly white men,” he wrote in 2006. “In a generation or less,” he says, “I suspect that the Obits pages (no doubt entirely digital by then) will be filled with stories of women and members of minorities who made contributions at a time when the world finally allowed them to.” Obituaries, in this view, are purely documentary efforts, untainted by bias. They seek to capture the accomplishments of the past without editorialization. As times change, so, too, will obituaries. Since 2006, McDonald and others on the Times obits desk have taken a more active role in bringing about that change. Case in point: “Overlooked,” which aims to document the losses that past generations of obituary writers deemed unmemorable, like Lilian Lindsay, the first British female dentist, and Alice Ball, a pioneering Black female chemist.

“Overlooked” and other projects like “Missing Them” at The City accomplish something automation and AI cannot: they recognize the political character of obituaries and seek to nudge memory work toward self-consciously reparative ends. They recognize how obituary pages have historically presented a skewed representation of America, leaving out “everyday people.” But they still produce obituaries in a highly standardized form, accepting that public memorial ought to look like a CV.

It is true that tools like Finding Words may help mourners write and place obituaries for their loved ones. And it is possible that this could lead to a greater representation of everyday people and gradually change the character of obituary pages. But obituaries should be more than just an account of the extent to which the deceased measured up to prevailing ideas of success.

LLMs like those behind Empathy’s “Finding Words” are likely to produce only pro forma lists of achievements and merits. If we automate obituaries for everyday people, for those unlikely to make a New York Times obituary section, we might reserve the dignity of human-generated memorialization for only the most accomplished and famous among us. The presence of the human hand is clear and comforting: someone took time to reflect on a life, to produce a coherent memory, and to communicate their human emotion for the purpose of dignifying and properly honoring a loved ones’ life. One need not spiritualize the process of writing an obituary to realize that more formulaic approaches to the process of grief and reflection that risk replicating our historical biases miss what makes the act of memorializing a human life significant.