The Art of Haunting

The email was innocuous enough. A woman wrote to me explaining how her partner had been talking about my work and how much he’d enjoyed it—in particular, he’d remembered me giving a TEDx talk about ghosts and hauntings. He’d tried to show her the video, but the two of them couldn’t find it online. Did I have a link I could send them? 

I wrote back confused; after thanking them for the kind words, I told them I’d never given a TEDx talk. I thought that would be the end of it. Instead, her partner wrote back to say that he was still looking for the video, and in the meantime he wanted to send a description of the TEDx talk I had given. Perhaps he thought that seeing the description of my talk would jog my memory, or perhaps he thought I was somehow lying. It was unclear. But he’d asked ChatGPT for the summary, and it had dutifully complied. It read:

“The Art of Haunting” by Colin Dickey is a compelling TEDx talk that delves into the realm of the supernatural through the lens of a skeptical journalist’s personal experiences. Dickey begins by recounting his journey into the world of paranormal investigation, initially driven by a desire to debunk myths and expose frauds. However, his perspective shifts dramatically when he finds himself confronted with a series of inexplicable events that challenge his rationality.

The crux of Dickey’s narrative revolves around his encounter with the ghost of an artist who, from beyond the grave, implores him to tell their story through a documentary. The artist’s persistent presence manifests in various unsettling ways, leaving Dickey feeling both haunted and intrigued. He describes visiting the hotel room where the artist died, immersing himself in the eerie atmosphere in an attempt to understand the spirit’s lingering presence. Dickey vividly recalls sitting in the bathtub, contemplating the final moments of the artist’s life, and experiencing a profound sense of connection to the supernatural realm.

One of the most poignant moments in Dickey’s tale involves his attempt to connect with an elderly woman who holds crucial information about the artist’s life. Tragically, before he can reach out to her, she passes away, leaving Dickey grappling with the realization that some mysteries may remain forever unsolved. Despite his skepticism, Dickey is forced to confront the possibility that there are forces beyond his comprehension at play.

Throughout his talk, Dickey intersperses his narrative with poignant reflections on the nature of belief, skepticism, and the human propensity for storytelling. He grapples with questions of identity, memory, and the thin veil between the tangible world and the realm of the supernatural. One particularly resonant quote from Dickey encapsulates the essence of his journey: “I’ve come to understand that skepticism is not the absence of belief, but rather the belief in a world where mysteries persist, even if we never fully unravel them.”

In the end, “The Art of Haunting” serves as a powerful testament to the complex interplay between reason and intuition, skepticism and belief. Through his captivating storytelling and introspective musings, Dickey invites audiences to consider the possibility that there may be more to the world than meets the eye, urging them to embrace the enigmatic beauty of life’s unexplained phenomena.

It was strange to be presented with this depiction of oneself that was so confident and yet so wrong. I am not a paranormal investigator, nor do I make it my business to expose frauds. I’ve never been visited by the ghost of an artist, and I’ve never been moved by any paranormal experience to make a documentary about it. But the guy who emailed me, at least, was satisfied that this was me. For him, the ChatGPT record was enough.

A bit frustrated, but trying to be gracious, I replied to affirm that no, I had never given a talk like this, that I had never given a TEDx talk of any kind, and that ChatGPT had invented this whole thing out of thin air. This is, after all, what the technology is designed to do: It bullshits (or “hallucinates,” to use the euphemism). The quotation about skepticism attributed to me, in fact, is a hallmark of Large Language Model (LLM) writing: Syntactically, it seems fine, until you realize that what I’m saying is that “skepticism” is itself “belief in a world where mysteries persist,” turning language on its head entirely and making me sound like a fool. 

As the weeks went on, this phantom talk began to gnaw at me, in particular this other Colin Dickey, this person who was not me but appeared vaguely similar to me. The notion of AI being able to replicate and distort the public perception of virtually anyone or anything has gone from being a dystopian technological fantasy to a reality so quickly that it already seems banal to mention it. In the sea of ersatz doubles, my own doppelgänger seemed hardly noteworthy. And yet, it was a reminder of where we’re going: As AI models corrupt more and more of the open web, realistic images and authoritative-sounding bullshit are only going to multiply. First interrupting the truth, then competing with the truth, and then, finally, through sheer volume, dominating and fully eradicating the truth.

A shift seems to be underway. For years, we’ve understood that humanoid robots and near-realistic digital animation can fall into the “uncanny valley,” a term coined by robotics engineer Masahiro Mori in 1970. Mori foresaw how, as machines became more lifelike, they would at a certain point become more, not less, unsettling, triggering the sense of something being off, even if one could not say exactly what that was.

Mori’s idea of the “uncanny” strongly resembles Freud’s concept by the same name, which, in turn, he formulated based partially on his colleague Otto Rank’s work on doppelgängers. In 1914, after seeing the silent horror film The Student of Prague, Rank had become fascinated by the figure of the double, producing a short survey on the topic that looked at both anthropology and literature. He concluded that the double was, in reality, the guilty conscience, a manifestation by a figure unable to “accept the responsibility for certain actions of his ego.” Freud then folded this into his classic 1919 essay, “The Uncanny,” subsuming the doppelgänger under this broader term, which he’d ultimately define as the return of that which was long repressed, coming to light abruptly in a guise both terrifying and familiar. The sense of the uncanny, exemplified in things like dolls, clowns, or haunted houses, is triggered by the fact that it is close to being familiar and comforting, but is still somehow off. 

But the last few years have revealed that the doppelgänger and the uncanny are further apart than Freud believed, and it seems as though we are rapidly leaving Mori’s uncanny valley and emerging onto the flat plane of the doppelgänger. With the uncanny, after all, you are overcome with a feeling that something is not quite right. But while you perceive your doppelgänger as also being wrong, the true terror comes from your awareness that no one else does. Balduin, the protagonist of The Student of Prague, is horrified to learn that his copy can intervene in his life in ways that the original will subsequently have to answer for. The fear of the doppelgänger is not due to its being uncanny, but that it’s not uncanny enough—that it’s a perfectly acceptable substitute as far as others are concerned.

What we’re learning about the world of LLMs and generative AI is that there are too many people who simply cannot tell the difference between the fake human and the real, and who accept chatbots as adequate replacements. One can see, then, why in many folk traditions, to catch a glimpse of one’s double (sometimes referred to as a “wraith” or a “fetch”) is believed to be a harbinger of one’s own death.

I began to fear that no matter how I felt about this other Colin Dickey, his work and ideas would soon enough replace mine. Last September, an AI-generated R&B artist got a $3 million record contract. Not long after, Warner Music Group signed a deal with AI-content mill Suno (after initially suing them), and soon, users will be able to ask generative AI to create content with the likeness and voice of established pop stars. We’ve already reached an age when AI doppelgängers have begun to compete with living musicians; surely creators of other artforms are not far behind.

This is the future that writer Italo Calvino imagined almost sixty years ago. In 1967, the Italian fabulist gave a talk envisioning a near future in which computers might produce literature of their own. Drawing on structuralist work that broke narratives down into functional segments, Calvino suggested that computers could, in principle, be made to generate their own stories. “Mankind,” he wrote, “is beginning to understand how to dismantle and reassemble the most complex and unpredictable of all its machines: language.” With such developments, he asked, “will we have a machine capable of replacing the poet and the author?”

The answer, he suggests, is yes. What’s more, a computer writing machine might be capable of replacing the very idea of the author itself. “The so-called personality of the writer exists within the very act of writing: It is the product and the instrument of the writing process,” he concludes. “A writing machine that has been fed an instruction appropriate to the case could also devise an exact and unmistakable ‘personality’ of an author.”

The ghost story will always, on some level, fail to capture the haunting, just as the algorithm will always fail to capture the life—both are anathema to the very existence of the unpredictable and the messy.

If writers themselves are nothing more than writing machines (“at least they are when things are going well”), why not replace them altogether? Perhaps to push back against the prevailing idea that art is the result of some ineffable, undefinable “inspiration,” some unquantifiable “genius” that is husbanded by rare artists, Calvino wants to strip bare the basic elements of storytelling and expose how easy they might be to replicate for a computer sophisticated enough to process them all. “The work will continue to be born, to be judged, to be destroyed or constantly renewed on contact with the eye of the reader. What will vanish is the figure of the author…that anachronistic personage, the bearer of messages, the director of consciences, the giver of lectures to cultural bodies.” Arriving the same year as Roland Barthes’s famous essay “The Death of the Author,” Calvino’s talk imagines an authorless machine capable of producing equally compelling literature, in language that seems entirely in line with the current hype surrounding generative AI’s ability to produce new writing.

What in 1967 felt perhaps far-fetched, nearly sixty years later feels more and more like a given. As a flood of AI slop is pushed out through various social-media channels, apps, and self-publishing companies, we’re seeing some of it start to break through. We appear to be on the cusp of a new landscape where human creators are shunted to the side in favor of the endless firehose of slurried content, some of which, through sheer, brute odds, arrives seemingly coherent, and perhaps even enjoyable.

But Calvino saw a countervailing force to this mathematical combinatory power. As the title of his talk, “Cybernetics and Ghosts” (the lead essay in his collection The Uses of Literature), suggests, he saw a binary at play in the creation of literature. Having prophesized the ability of computers to construct literature, gathering up all that’s already been said, analyzing patterns and syntax and not only recreating it but authoring new permutations and new literature, he then shifts to what has yet to be mentioned: the mythic, the unconscious, the unsayable, the repressed—in short, that which will always escape the ever-expanding grasp of the LLM algorithm, the thing which can be felt but never directly apprehended—the ghost.

“The power of modern literature,” Calvino explains in the second half of his essay, “lies in its willingness to give a voice to what has remained unexpressed in the social or individual unconscious: This is the gauntlet it throws down time and again.” Calvino’s point, as I see it now, is that breaking down existing language, analyzing it and taxonomizing it, and then recombining it to form new stories—all this is fine for emails and things no one wants to read anyway, but it’s inherently regressive and misses what makes literature vibrant. Or, as he puts it: “The more enlightened our houses are, the more their walls ooze ghosts.”

Rereading Calvino’s description of this binary—the computer and the ghost that haunts it—I came to understand the problem with my doppelgänger’s TEDx talk: He’d misunderstood his own title. He had no notion of art, and no real notion of haunting.

I began seriously cataloging ghost stories over ten years ago for my book Ghostland: An American History in Haunted Places. As I got further into the project, I started to notice certain motifs appearing over and over again in a variety of places. For example: The Wayside Inn in Sudbury, Massachusetts, is haunted by a woman whose lover died at sea, and she waits endlessly, it seems, for his return. She is not unlike the ghost who haunts Senior Hall in Stephens College, in Columbia, Missouri, who also awaits her lover. As does the Lady in Blue, who haunts the Palace Hotel in Port Townsend, Washington. Meanwhile, every Civil War battlefield is haunted by some soldier or other who still hears the spectral bugle call and stands ready to fight even after death. Every old prison is haunted by the ghost of an inmate who died in a particularly brutal manner. And in places as disparate as Savannah, Georgia; Moundsville, West Virginia; and Los Angeles, California, ghosts of murder victims whose killers were never found still linger, demanding justice. 

Over time, these ghost stories have a tendency to become stock narratives, the same details constantly repeated, their edges smoothed off, their specificity lost to time. The purpose of such stories becomes less about exploring the paranormal and more about reaffirming some basic cultural beliefs: that true love transcends death; that a soldier’s duty persists past the final sacrifice; that justice denied in this world can be redeemed in the next. 

What I’ve found is that the stories we tell about ghosts are usually narratively unified and complete, and that their goal (other than to give us a quick jolt of terror) is to validate truths we learned in childhood. Honed and hallowed through repetition, they are predictable, stable. They reassure; they tell us that things we were taught are true.

Meanwhile, the stories I actually hear from individuals who’ve had some paranormal experience often take a very different form. When people tell me about what they’ve seen or felt, the stories are more amorphous, more ambiguous. Someone will describe a radio suddenly switching on in the next room, or their taking a turn down an alley only to be met by a floating blue figure. Sometimes it’s just a feeling or a noise. A strange light in the sky. Rarely does someone have a full and complete story. Most often, all they have are fragments.

For me, this is the difference between the “ghost story” and the “haunting.” The ghost story is a narrative, with a complete arc and a sense of resolution. And it always has, either explicitly or implicitly, some kind of moral message. The haunting, on the other hand, as actually experienced by people, is often elusive, inconclusive. The world of hauntings—of blurs and unexplained presences, of eerie absences, strange smells, inexplicable sensations—is one where we are confronted by what we don’t know, what’s still out there beyond our ken. It is not a place of familiar tropes and themes. It is a place without any clear meaning whatsoever.

What bothered me about my double’s story is that it was complete, with a simple moral message at the end. He started out a skeptic, encountered a ghost, and became a believer. The ghost is undeniable, fully realized, and comes bearing a tragic story. By the end, a lesson has been learned. This Colin Dickey was uninterested in ambiguity or matters left unresolved—or, if he was, it was only to vindicate an overarching truth.

I can’t entirely fault him; it’s how he was built. OpenAI, the company behind ChatGPT, has acknowledged that LLMs will always, by design, invent facts and “hallucinate,” in part because they lack the ability to admit that there are things that they simply don’t know. As Neil Shah, VP for research at Counterpoint Technologies, explained recently, the problem with this technology is that, unlike “human intelligence, it lacks the humility to acknowledge uncertainty. When unsure, it doesn’t defer to deeper research or human oversight; instead, it often presents estimates as facts.” This is a technology that is hostile to ambiguity, to the unknown and the unknowable.

My double, possessed of this same hostility to ambiguity, has reduced life to a satisfying and stable narrative. Unlike me, he believes that our attitudes toward ghosts can be divided into “skepticism” and “belief,” and that there is no middle ground between these two poles. His writing, therefore, tends to reduce any apparent complexity down to its simplest forms.

As for me, I’m skeptical of the idea that one’s life (or death) can be laid out so neatly. I’ve long been somewhat distrustful of the kind of storytelling that always leaves out the inconvenient, that sands off the edges and fudges the details in order to create something narratively satisfying. I recognize the seduction inherent in packaging the detritus of one’s life into a pat anecdote where each beat is known fully in advance—but perhaps the reason I’ve never actually been invited to give a TEDx talk is because I have no interest in condensing the story of my life into such a takeaway.

Literature, at its best, produces a certain kind of “shock,” and it’s a shock, Calvino reminds us, “that occurs only if the writing machine is surrounded by the hidden ghosts of the individual and of his society.” The gauntlet thrown down in any attempt to write is to give voice not to the familiar and commonplace, but to the strange, the uncanny, the ghostly that interrupts waking life and offers up what might not otherwise be expressed. The ghost story, in other words, will always, on some level, fail to capture the haunting, just as the algorithm will always fail to capture the life—both are anathema to the very existence of the unpredictable and the messy. The goal of good work, rather, is not to be reassuring, not merely to affirm platitudes, but to push the boundaries of our understanding into the unknown, to force a confrontation with the unexplained presence, the mysterious other. To seek that which cuts against the banality of the world, and the writing that chases this without ever fully catching it. 

Let me put it another way. “The Art of Haunting” by Colin Dickey is an essay that delves into the question of how to tell a ghost story in the twenty-first century, as Dickey asks what it means “to haunt.” The essay ends with Dickey describing how he became haunted by a number of mostly nameless artists on a trip to the West Virginia Penitentiary, in the town of Moundsville—often cited as the most haunted prison in the country. A massive, fortress-like building along the Ohio River, it was built in 1866, its stone walls quarried by its own inmates. Originally meant to hold 650 inmates, it quickly reached capacity and over the next hundred years became overcrowded and fetid. Nearly a thousand inmates died within its walls through the years, dozens of them homicides. 

Strict order was maintained by the guards, often brutally. Monotony and predictability become, in their own way, part of the punishment of prison life. In prison life, everything that will come before you derives from everything that’s already happened. There’s little chance for diversity, improvisation, or chance.

And yet, over time, the walls began to ooze ghosts. After the prison’s closure in 1995, it was reopened for tours, a seasonal haunted house, and ghost hunts. Visitors have since claimed to experience all manner of paranormal activity, including sightings of the ghosts of men killed here. The stories themselves are uninteresting and largely mirror the kinds of ghost stories you hear in other haunted prisons—Eastern State in Philadelphia, the Ohio State Reformatory, or Alcatraz: Evil men get punished, in this life and then again in the next. You’ve heard it a dozen times before.

Dickey toured the penitentiary in 2015, and while he admits he never saw anything supernatural, as he explains in his essay, he became haunted by something else. At some point during the prison’s history, staff began allowing prisoners to decorate the walls, and now its interiors are covered in murals. During the tour, he saw amateur depictions of cartoon characters, bucolic scenes of waterfalls and forests, a leaping bobcat and a sixteen-wheeler stretching the length of the cafeteria wall, and dozens of other doodles and sketches. A few of the artists’ names are known, but most have been lost. Dickey recalls being struck by these images, which were amateurish at best, hardly rising to the level of outsider or folk art. And yet, he confesses, he could never quite get these images out of his mind, even after writing about them. Like a ghostly presence—a whisper in the dark, or a sudden chill—the art suggests the shape of a person’s life, a story that can never be fully told.

As long as the prison stands, it will be a strange canvas that bears witness to the souls that passed through it. At the end of his essay, Dickey recounts how he knew of the building’s haunted reputation when he made the trip, but had no idea about the art, a completely unexpected portal into the sublime.

These unknown artists…why were they here? Were they inhuman, antisocial criminals who deserved to be removed from society? Were they minor thieves or thugs, people who’d screwed up but weren’t beyond rehabilitation? Were they innocent men? Or perhaps men who’d long served whatever time they should have but lingered in the limbo of a broken carceral system?

Dickey confesses he’ll never know the answers to these questions. Perhaps he doesn’t want to. Would it matter if one was haunted by the work of someone immoral and cruel, instead of one unjustly condemned? Does our reaction to art—or to the supernatural—depend on knowing every detail? Or is it in fact the case, Dickey wonders, that what truly haunts is the thing that doesn’t meet us all the way, that refuses any resolution or the formulaic narrative you might try to shoehorn it into, that instead forces on us a confrontation with the ambiguous and the unknown? The thing that hovers in a dark hallway, asking questions that can’t be answered?

You stand before the work of art. The artist, long gone, is reaching out to you from across some awesome distance, imploring something of you. Is this presence malevolent? Does it speak of some lost tragedy? Impossible to know for sure. All you can say is that there’s something here.

What happens next?