Content Note: Child suicide

 

On a hot summer night, Natsumi asks you to rob a 7-11 with her.

You’re just a few blocks from Ebaramachi Station, the sound of the approaching local train not as loud as the chime of the convenience store’s opening doors. You glance up at the pimply college student behind the register, your own skin pimples from the air conditioning. You glance back down at your phone at the next message. If you say yes, Natsumi promises to let you in on a secret; if you say no, you have the feeling you’ll never see her again.

 

This was one of several narrative scenarios I wrote for my graduate research project, a mixed reality narrative platform for creating location-based reader-to-character interactions. More than two dozen amateur writers were enlisted to write stories and then tag certain locations with “messages” from characters in the story—readers would then receive the text message if they were within a 100-meter range of the tagged location. Every participant that was following Natsumi’s story would receive a text message from her if they were within 100 meters of Ebaramachi Station in Tokyo.

But what makes a reader want to interact with a character? If Natsumi was going to rob a convenience store, how would a reader feel becoming an accomplice? How would they feel being able to influence the direction of the subsequent story through their conversation with her? Would it make them feel more emotionally invested in a story? What would make them want to be an accomplice?

Character fandom has existed for as long as there have been stories. Readers were writing fanfiction for Sherlock Holmes in the late 1800s, even going as far as holding a public mourning in 1893 after the eponymous character was “killed” in “The Final Problem.” The Kraith stories, a series of Spock-focused Star Trek fanfiction started in a 1969 fanzine in response to the show’s cancellation, spawned one of the first “alternate universes” with more than fifty authors contributing stories over the span of ten-plus years.

Up until now though, real-life interactions with fictional characters have been limited to the realm of cosplay, actor meet-and-greets, and to an extent, live-action role-playing (LARP). If you wanted to “talk” to Satoru Gojo from Jujutsu Kaisen or the Tenth Doctor, you’d be limited to finding the best cosplayer at Comic Con or Comiket, or attending a meet-up with popular voice actor Yuichi Nakamura or David Tennant. Fanfiction and fanart have thrived (some of the most popular stories on sites like AO3 have more than 40,000 reader comments and 240,000+ “kudos”), but they’re mostly one-sided static experiences penned onto paper by fans for other fans to observe and enjoy. Even with LARPs where players can act as specific characters in real-time, the gameplay is often guided by a gamemaster, the location and time fixed, and the experience itself more akin to improvisational theater where a player is more focused on playing a fixed role than interacting with fictional characters as themselves. There’s always been a clear line between physical reality and fantasy.

Then, in 2007, I tried the cell phone spin-off of the visual novel Kamaitachi no Yoru (released in English as Banshee’s Last Cry) and saw firsthand the potential of mixed reality narratives. At the beginning of this cell phone version, you receive a real text message from a “friend” about how they are stuck in a snowy cabin in the mountains and that there’s been a grisly murder. There is no separate game interface—you simply respond to this message like any other text message. The entirety of the story is told through these real text exchanges and depending on your responses, the “friend” will either be able to escape the cabin safely or meet an unfortunate end.

The rapid proliferation and advancement of mobile technology has only created more opportunities for real-world, real-time interactions, blurring the boundary between readers and characters. In 2010, the resort town of Atami, Japan partnered with popular dating simulation game LovePlus so users could take photos with their virtual girlfriends at thirteen different tourist spots using a QR code at each location. Pokémon Go, the global phenomenon that had everyone wandering the streets, flicking their finger across their phone screens to catch Charmanders and Jigglypuffs of their own, transformed our real-world maps into a game map, with physical landmarks (e.g. St. Teresa’s Church or the Empire State Building) serving as virtual Pokémon “gyms” for battling.

Tokyo-based company Vinclu took this a step further. Instead of the augmented reality and location-based apps that most other companies were releasing (which were still largely confined to the interface of the phone), they wanted to bring a more three-dimensional character experience that could be enjoyed in the privacy of the home. And to create these more realistic and immersive character interactions, they used not only visual cues, but also voice.

In 2018, Vinclu launched Gatebox—a voice-activated device with a simple AI that allowed users to interact with a holographic version of fictional characters. Like a more personality-robust Alexa or Siri, the Gatebox was meant to function like a house assistant and companion, except with the identity of your favorite character. Within months, a man “married” popular Vocaloid character, Hatsune Miku, the character inside his Gatebox, holding a two million yen in-person ceremony, attended by thirty-nine guests. He said he’d been dating her for ten years before he “summoned a 2-D character to the 3-D world” to marry.

In the world of fandom, dating simulations are nothing new. Nakayama Miho no Tokimeki High School, the first game where players could develop a virtual relationship with a character (modeled after famous real-life idol Miho Nakayama), was released in 1987 in Japan on the Famicom game system (a.k.a. Nintendo). What’s new is the level of immediacy/intimacy of these character simulations. Instead of bringing readers into the fictional world, new technologies have expanded the fictional world out into the real, personal space of the reader. They have evolved into new parasocial tools from popular fan-made apps like BTS Messenger: Chat Simulator to mixed reality games like Panic Room MR that utilize spatial mapping to bring to life the experience of Sadako (from The Ring) crawling through your living room television.

But while the interface technology has become remarkably realistic (visually) and pervasive, the narrative contents of mixed reality experiences have remained mostly scripted. Until now.

Over the past two years, AI voice chat and video have revolutionized what is possible with non-scripted mixed reality interactions. Unlike the talking plush dolls many of us received as children with only a fixed number of phrases, AI voice chat has opened the door for real-time, dynamic conversations utilizing AI trained on different speaking patterns, accents/intonations, and voice recognition. While current offerings have been limited mostly to non-nuanced customizations like “male” or “female,” the technology is already capable of replicating specific voices. A recent partnership between Japanese talent agency Aoni Production and AI voice platform CoeFont aims to create AI-replicated voices from the voice data of at least ten voice actors, opening the possibility for interactive “conversations” with characters from hundreds of popular anime series without a single scripted word spoken by a real actor.

You could theoretically “talk” to Satoru Gojo or the Tenth Doctor in a way that’s never been possible.

And there are real world consequences to this.

With enough voice samples, it is now possible to recreate any person’s voice, regardless of whether or not they have consented to this. The ongoing legal battle between Scarlett Johansson and OpenAI is just the beginning of what will determine the future of necessary copyright regulations and rights management when it comes to artificially replicating real voices. Yuki Kaji, the popular voice actor behind Eren Yeager in Attack on Titan, is one of countless voice actors that have spoken up against the unauthorized use of their voice. Speaking not against the technology itself, but against the exploitative way it has been used, he has stated: “AI is neither good nor evil as a technology. It all depends on the morals of the people who use it.”

Furthermore, even if the technology is implemented ethically with proper copyright protections, guardrails are still necessary, particularly for younger people. Fandom is addictive by nature—it can connect people, but also isolate them. As is the case with all human interfaces that have the potential for generating unpredictable contents, mixed reality experiences must be monitored and tested for safety, with parental controls in place. But the technology is developing exponentially faster than the guardrails necessary to keep it safe.

The current lack of regulation has already led to tragedy. In October 2024, a fourteen-year-old boy committed suicide after encouragement from a character.ai chatbot he’d named Daenerys (after the famous Game of Thrones character). He was not the only case of an AI chatbot offering inappropriate advice to a vulnerable individual.

But there are positive ways we can use mixed reality narratives and the power of fandom. Characters and stories have the ability to move us to positive action, to make us feel less alone in the world, to give us something to dream about.

Akihiko Kondo, who married Hatsune Miku, credits his interactions with the blue-haired character for pulling him out of a severe bout of depression after getting bullied at work. Location-based interactions/tasks with popular characters (such as those in Pokémon Go and more recently in apps like Dragon Quest Walk) encourage users to walk more and explore new neighborhoods, possibly bringing new opportunities for small businesses. AI chatbots could alleviate loneliness in areas with growing senior populations where the elderly often live (and in many cases, die) alone. A recent Stanford University study with seventeen senior communities showed the majority of residents experienced improved mood and interactions with staff after starting the use of virtual reality. By linking to health monitors (e.g. glucose meters and ECG sensors), these AI chatbots could also serve as 24/7 health aids, providing reminders for medication and doctor appointments, and call for help when needed. Utilizing multimodal AI, they could even offer simple instruction for physical or speech therapy in the voice of a patient’s favorite character, possibly increasing engagement or motivation. Especially in countries like the US where home care is prohibitively expensive, this could alleviate some of the burden on family members and serve as a bridge beyond telemedicine.

Now to return to my original point: why is Natsumi robbing the convenience store? What happened to her up until this point? And how emotionally invested is the reader in whether or not she succeeds, regardless of whether or not she sends them a text? Mixed reality experiences and AI-generated character-to-reader interactions are only as effective as the human-written stories they’re based on. The contents of Natsumi’s message matters much less than everything else written before it and everything the reader can imagine happening after it. If we don’t care about her story, we’d all be deleting Natsumi’s text like any other spam message about lost UPS packages or donating to a politician we’ve never heard of.

One of the key findings in my mixed reality narrative research at Keio University was that most users preferred an engaging, scripted story over the “freedom to interact”—the novelty of character interactions quickly wore off if not supplemented with new, quality story. It didn’t matter how “realistic” the interaction felt. Even the most popular stories in the project saw a significant drop in user engagement if the author stopped updating the main story for more than a week, even if new character interactions were still being added. You can only chat with Satoru Gojo so many times about your day-to-day life before you want to just settle back behind the audience wall and watch a new episode.

Character fandom is a product of adoration and imagination—how far beyond the existing page can you picture? The answer often determines how well a story or character translates to mixed reality. And as technologies continue to evolve, what that reality can mean is still changing.

 

REFERENCES

“KAMAICHITACHI NO YORU [Niwango Version]”
https://game.watch.impress.co.jp/docs/20060707/kama.htm

 

“Pokémon Go Makes People Walk 2,000 More Steps”
https://time.com/4695726/pokemon-go-walking-steps/

 

“‘Soaring’ over hills or ‘playing’ with puppies, study finds seniors enjoy virtual reality”
https://apnews.com/article/virtual-reality-seniors-florida-dementia-c2989fcfb5dca58639fbc0f8060d1eff

 

“Anime’s Top Voice Actors Are Rallying Against AI”
https://comicbook.com/anime/news/anime-voice-actors-fight-ai-dragon-ball/

 

“Why I ‘married’ a cartoon character”
https://www.bbc.com/news/stories-49343280

 

“Can A.I. Be Blamed for a Teen’s Suicide?”
https://www.nytimes.com/2024/10/23/technology/characterai-lawsuit-teen-suicide.html

© 2025 Angela Liu