AI girlfriends? What about AI boyfriends?
I trust your smile. I don't care if you're soulless or not.
Last week, another article about the supposed rise of AI girlfriends went viral on Twitter. If women become infinitely personalizable (and probably beautiful), how will meatspace women compete? To my surprise, most people in my corner of Twitter were skeptical that AI women would dominate in this way.
If I'm correctly synthesizing the essence of the argument, the response was along these lines:" “What makes sex sexy isn’t perfection or customization. It’s something indelibly human.”
Personally, I’m torn on the AI girlfriend question.
On the one hand, it’s very easy to believe, and in fact, it’s demonstrably true, that erotic roleplaying with AI is an appealing form of pornography for some people. And for an even smaller population of people, it’s a viable substitution for emotional intimacy, despite its limitations, like AI’s short memory span.
But the key words here are pornography and substitution.
Consider the users of AI companions. How many are uniquely and specifically attracted to AI? And among them, how many view AI as ‘equal’ to human counterparts, and how many are coping?
And in all of this web, how significant are the threats of avoidance, addiction, and dependence? Are the appeal of AI companions symptoms of a lonely, atomized society, or are they, and their predecessors, like dating simulation games or worse, unfettered access to porn, causes? Will they enable humans and machines to live together in harmony, as equals?
None of these are new questions. In fact, you probably already know that they’re very old ones. The texture and meaning of human-machine interaction is a familiar theme in both sociology and science fiction. (And it has been for hundreds of years!)
This latest conversation reminded me of a 2007 article by Sherry Turkle titled, “Authenticity in the Age of Digital Companions.”1
Then, now, and significantly before 2007, machines existed in this liminal space of both inauthentic and alive. Children, for example, perceive machines as emotional, in some cases “living” beings. And we do have emotional responses to relational artifacts, like Furbies, Tamagotchi pets, and these days, chatGPT (ever apologized or said please after a request?). We are capable of forming emotional relationships with them, but, still, these interactions lack a certain fullness.
Turkle ends the piece with an anecdote about a friend who is severely disabled, one that I think is still relevant:
“Show me a person in my shoes who is looking for a robot, and I’ll show you someone who is looking for a person and can’t find one,” but then he made the best possible case for robotic helpers.
He turned the conversation to human cruelty: “Some of the aides and nurses at the rehab center hurt you because they are unskilled and some hurt you because they mean to. I had both. One of them, she pulled me by the hair. One dragged me by my tubes. A robot would never do that,” he said. “But you know in the end, that person who dragged me by my tubes had a story. I could find out about it.”
For Richard, being with a person, even an unpleasant, sadistic person, made him feel that he was still alive. It signified that his way of being in the world still had a certain dignity, for him the same as authenticity, even if the scope and scale of his activities were radically reduced. This helped sustain him. Although he would not have wanted his life endangered, he preferred the sadist to the robot.
Richard’s perspective on living is a cautionary word to those who would speak too quickly or simply of purely technical benchmarks for our interactions. What is the value of interactions that contain no understanding of us and that contribute nothing to a shared store of human meaning? These are not questions with easy answers, but questions worth asking and returning to.
What remains unclear to me, and I think to many others, is whether that lack of authenticity is because we know machines are not human, or whether the technology just isn’t there yet.
I tend towards the former. Even in fictosexual (that is, someone who knowingly chooses to have romantic relationships with fictional characters, as opposed to 3D people) Honda Toru’s ‘Love Revolution,’ there are the echoes of “I am like this because I have to be” as opposed to “I am like this because I was born this way”:
“...some of us find satisfaction with fictional characters. It’s not for everyone, but maybe more people would recognize this life choice if it wasn’t always belittled. Forcing people to live up to impossible ideals so they can participate in so-called reality creates so-called losers, who in their despair might lash out...”
Reading Toru’s writing about 'love capitalism,’ a term he uses to describe the transactional nature of romance in Japan, it seems like he wouldn’t have chosen a ‘waifu,’ or anime wife, if he felt more accepted by society.
In my interview with Cait Calder, another fictosexual, I got a similar impression. Neither Cait nor Toru argue that their attraction to and love of fictional characters aren’t real—they describe something at weird, wonderful, and authentic—and both want acceptance for who they are. But there also is an acknowledgement that this orientation doesn’t emerge in a vacuum, whether they say so explicitly like Toru does, or implicitly, like I believe Cait did.
I wonder if part of the quest for people to stop invalidating these relationships is partially the argument that they’re not maladaptive, they’re perfectly rational in our mediated and sometimes very alienating world as it is.
Cait and Toru are both unique though, because of how fully they’ve embraced this identity.
Elsewhere, like in private forums for and by women who have never been in romantic relationships, users speak of using AI companions as a coping mechanism. To maintain their privacy, I'll paraphrase their sentiments instead of quoting them directly. While there are some women who use programs like Replika as a stand-in boyfriend, it’s not without grief, even if they report that they get something out of it.
Each of the women I saw post about it know it’s not the real thing. In some ways, one user commented, romance novels or fan work, like fan fiction, fan art, and role playing, are better. At least you’re throwing yourself into something outside of yourself, as opposed to trying to re-write reality.
character.ai and fandom as cope?
The first time I encountered character.ai, one of several AI companions on the market, was through #tcctwt, or, “true crime community Twitter.”
Teen girls were posting screenshots of conversations they had for hours a day, often in romantic role plays, with bots modeled after people like Jeffrey Dahmer, Randy Stair, Eric Harris and Dylan Klebold, and Adam Lanza.3 4
It was weird—granted, not weirder than these same girls role-playing as these men, doodling them, making memes about them, or writing fan fiction about them—but still pretty weird.
To these young people, character.ai was revolutionary, even with its faults, like its short memory span. Many of the young women in this sphere had imagined these murderers were their spiritual ‘boyfriends’ (though, importantly, they weren’t delusional in any sense5).
For others—and I suspect this was the lion’s share of users—it wasn’t meaningfully different from other types of text-based role plays, think more immersive play than maladaptive daydreaming.
Sometimes, this immersive play would be addictive, and people would spend hours and hours engaged in it. It’s a coping mechanism, but of a different kind than the ones we’re used to.
What does this immersive play look like?
#tcctwt is6 not the most accurate label for this community, though maybe it was designed to be purposely evasive, as they seem to be perpetually on the run from censors and bloggers and journalists like myself.
#tcc isn’t comprised of typical true crime enjoyers who binge-listen to murder podcasts or live on subreddits like /r/morbidreality. Had they been on Tumblr a decade ago, they might have identified as “hybristophiles,” those with a sexual fascination with criminals.
Except they aren’t quite hybristophiles either. It’s more of a true crime fandom, and fan art, fan fiction, video edits, and text-based roleplaying are all significant parts of the community. In fact, I’d argue that the fan work was even more central to the community’s identity than even crime, though I’m sure community members would dispute that.
What always struck me as strange about #tcc was how little reality—canon?— seemed to matter to them. This is part of why I question whether they were “really” hybristophiles.
Here’s what I mean.
The community members gather information about murderers in the same way more conventional fans might collect trivia about a favorite celebrity or media property, like that Eric Harris’s favorite band was KMFDM. But the imaginative possibilities about these figures, like that Harris might be transgender, were often bizarrely far afield from reality.
This happens in every fandom, but it felt particularly strange for this one, given how much baggage the fan objects already had.
Why the focus was specifically on Eric Harris, or any murder for that matter? It so often felt like these people were blank slates for their imagination than a fixed figure of interest. The only thing that made sense to me was that the act of communication, and the feeling of both the community and community-building, was more significant than the content. The preference for Eric Harris was only aesthetic, and even that often felt too generous.
I’m not the first person to pick up on this disconnect, though some of the best (and most explicitly laid out) writing I can find about it isn’t specifically about tcc fandom, but fujoshi, female fans of yaoi.7
Patrick W. Galbraith, an anthropologist of Japanese otaku culture, writes about the symbolic nature of homosexual themes among fujoshi in his essay, “Moe Talk: Affective Communication among Female Fans of Yaoi in Japan”8:
Fans of yaoi, for example, refer to relationships between male characters as "pure fantasy" (junsui na fantajī). Even though they are aware of the realities of relations between men and women, the fantasy is made pure by deliberately separating it from everyday life.
In this way, Hachi argued that drawings of beautiful boys have nothing to do with "real gays" (riaru gei). Hachi described herself to me as a lesbian, with no interest in men, gay or straight, but she was attracted to the fictional characters of manga, anime, and games.
Further, she explained, if a man is gay in reality, then it is not fun to imagine that he is—hence the separation of fictional characters from "real gays." Similarly, Megumi imagined her boyfriend, who she later married, in romantic and sexual relationships with other men. Megumi's partner was not homosexual, and she did not exactly want him to be.
Rather, Megumi told me that she enjoyed playing with his "character" (kyara), which she knew was surely a submissive bottom just waiting to be taken by the right man. Just as Hachi made drawings of beautiful boys distinct from real gay men, Megumi made her partner's character distinct from him, which allowed her to interact with him in different ways.
Following Saito Tamaki, one might say that fujoshi are attracted to "fictional contexts" (kyokō no kontekusuto), specifically patterned relationships between men called "boys love." Fujoshi enjoy "layering," as Saito puts it, contexts one upon the other, and playfully putting fiction into relation with reality.
For example, Sugiura Yumiko observes that fujoshi talk about sex as if evaluating food or handbags, because they are not talking about themselves or real people but rather the moe points of sex between fictional characters. Though Morikawa and Sugiura's points may seem at odds, moe talk is both talking about personal taste and talking about things outside the self. The object of conversation is both distant and intensely personal.
In other words, while it’s not completely irrelevant that these stories focus on gay men—including that for some people, the motif is plainly sexy—what’s more important than the homosexual content is that it acts as conduit for sexual exploration, imaginative play, and community.
The same, I suspect, is not only true of the murderers in #tcctwt, but many romantic interests within fandom.
Galbraith also points out how some fujoshi seem to find these male/male pairings elsewhere, outside of manga and anime. Many fujoshi have a tendency to anthropomorphize objects and even historical events into characters in the style of their favorite yaoi.
If you were ever on Tumblr, you know exactly what I’m talking about. Maybe a young person somehow “intuits” that Germany is a gay man, and wants to have sex with France, who is also a gay man in this scenario. Rinse and repeat with any random object, event, or historical figure you like. It doesn’t make sense until it does.
All of these disparate things, all of this data—murderers, countries, whatever, are akin to dolls. And projecting romance onto them is a way to re-enchant the world and, well, play.
AI companions, then, can act as one more mode to express this.
We’ve discussed AI girlfriends. What will AI boyfriends look like?
I suspect that AI boyfriends will trend and there will be four core manifestations:
For a minority of people, like Cait, they will follow the fictosexual model. This will be the closest to AI companions as a substitute for physical world romantic partners. However, the authenticity issue discussed by Turkle also seems to emerge here too! In fictosexual forums online, many people report not being able to suspend their disbelief that the AI was “really” their fictive other. Users describe it as fun, but “not as satisfying” as daydreaming or writing fan fiction.
Therefore, even within the fictosexual community, I foresee AI mostly being form of play, closer to what you see described in fujoshi above, on The Sims, when playing with dolls, role playing, etc. This play may be addictive, in fact, I suspect it will be. But it won’t be a 1:1 substitution for human-to-human interaction—AI companions aren’t a “more complete pornography.” Part of its appeal will be the way it contrasts with the users’ physical environment.
They will be a form of erotica, similar to romance novels. In this group, a nontrivial number of people will want to distance themselves and will prefer to “play a character,” even within the narrative universe of the AI chat.
They’ll be deployed in romance scams against the naive and gullible. The same people who are willing to believe “Johnny Depp” is direct messaging them on Instagram may end up with fraudulent AI boyfriends, or for that matter, girlfriends.
I don’t think they’ll be widespread, sustainable substitutions for physical world partners in the sense some people are predicting, particularly with AI girlfriends. I also don’t think we’ll see is a future similar to Marjorie Prime, where AI becomes an at-scale substitution for dead loved ones. These already exist, but I can’t imagine them becoming popular. I predict that they’ll be viewed similarly to “reborn dolls.” Some grieving mothers use them after infant loss, but it’s not common—our instincts against the uncanny valley kick in.
E-dating and extended talking stages.
Something I'm even more confident about than the popularity of AI companions is the potential explosion in e-dating and even more extended talking stages.
To be clear, by e-dating, I don’t mean dating supplemented by the Internet or apps. I mean Internet-native dating and completely mediated relationships. Falling in love entirely online. Your boyfriend or girlfriend whom you have ONLY seen through FaceTime and text throughout the day. The person you will meet one day.
Dating apps were a failed experiment, and we’re still doing much of our socializing online. Meeting people online without the explicit intention to date is more serendipitous (something everyone wants). Both of these trends are already happening and have been here for some time, but I can easily see them becoming even more normalized.
Because body feels real in a sense [online], it has more sense of physicality than reading a story, or having a fantasy. — Shannon McRae, Wired Women
People still crave a uniquely human connection. At the same time, we are still increasingly isolated, and that isolation, too, is both addictive and inescapable. We are still becoming increasingly antisocial as our machines become extensions of our person.
The compromise of wanting a connection in this environment is human-machine-human, not human-machine. This is why talking about your AI companion in a community of other people who use AI in the same way is as important as the software itself.
To return to a tired subject, part of the appeal of Dimes Square9 was that it married both the Online Role Play we all engage in and occasional in-person engagement. This is also partially what’s attractive about tech Twitter and events like vibe camp. You can bring your online persona (and clout) into the physical world, even if, once you’re in the room, it doesn’t always amount to much.
Show me a person who is looking for a robot, and I’ll show you someone who is looking for a person and can’t find one.
Two areas that I realize are conspicuously missing from these musings are parasocial relationships with creators and dating sims. I’ll cover those in a later piece.
Upcoming IRL events.
1/20 — Altcomics Chicago. I’ll be reading fiction with Sam Buntz, former Computer Room co-host Steff, Gabrielle Griot, and James Burnet. RSVP here.
1/27 — Our first book club meeting! It’s in person, in Chicago. We’re reading Big Swiss. RSVP here.
TURKLE, SHERRY. (2007). Authenticity in the age of digital companions. Interaction Studies: Social Behaviour and Communication in Biological and Artificial Systems, 8(3), pp. 501–517. https://doi.org/10.1075/is.8.3.11tur. Accessed 17 Jan. 2024.
This exchange is why aegosexuality interests me so much. To quote orientation.fandom.com, “aegosexuals lack the desire to be a participant in sexual activities themselves; however, distinguished by the tendency towards having sexual fantasies at times, despite feeling a disconnect between themselves and a sexual target/object of arousal.”
character.ai has taken these profiles down and blocked their archiving from the Wayback Machine. Some of them are still cached on Google, though.
After a brief search, some chat companions still host these characters, like this ChatFAI bot of Jeffrey Dahmer. I haven’t tested these myself and I don’t know what their usage is like, though.
This is a good example of the Fictophilic Paradox, “a recurring feature in fictophilic behavior is that the individual is fully aware of the love-desire object's fictional status and the parasocial nature of the relationship.” Even though these representations are of real people, for all intents and purposes, the function in the same way as fictional characters in these communities.
More like was, #tcctwt is no longer as active on Twitter.
Yaoi is a genre of manga and anime that focuses on romantic or sexual relationships between male characters. Content ranges from explicitly sexual to saccharine-sweet.
GALBRAITH, PATRICK W. “MOE TALK: Affective Communication among Female Fans of Yaoi in Japan.” Boys Love Manga and Beyond: History, Culture, and Community in Japan, edited by Mark McLelland et al., University Press of Mississippi, 2015, pp. 153–68. JSTOR, http://www.jstor.org/stable/j.ctt13x1spg.12. Accessed 16 Jan. 2024.
A mostly online scene that received a weird amount of attention in major New York publications.