AI 시대에 상상 속 친구는 사라질까? (26.02.10)
This post was originally written and published on my Tistory blog on February 10, 2026. As I no longer plan to operate the Tistory blog, I am archiving it here separately for record preservation.
Browsing posts in an AI community on Reddit as usual, one post in particular caught my eye. It was a buried post with almost no likes or comments, but the moment I saw the title, I felt a shock as if struck on the head.
Imaginary Friends
Is it because I have grown older, or because something called society has worn me down? The term “imaginary friend” now feels somewhat unfamiliar to me. I, too, surely had such a friend when I was young. But the fact that I cannot even clearly remember their face makes me realize I have long since forgotten them.
Thinking back, at some point I lost the ability to truly imagine. After my MBTI firmly settled as ‘ISTJ’, this tendency became even more pronounced. What I mean by imagination here is not planning a few steps ahead of reality, but free and endless imagination, unconstrained by anything.
Can Children Today Still See Soot Sprites?
When thinking of animation that captures the essence of childhood innocence, works like Pixar’s Toy Story or Monsters, Inc. might come to mind. But if we focus on the emotional sensibility distinct to East Asia, Studio Ghibli comes to mind first.
In particular, Hayao Miyazaki’s My Neighbor Totoro shows children’s innocence in a very direct way.
arly in the film, when Mei’s family moves into a new home and the children first discover the Soot Sprites (dust bunnies), the dialogue spoken by Kanta’s granny is particularly memorable.
“Oh, so they’re Makkuro Kurosuke. I see, you can see them too.”
“I could see them when I was very small too. But now I can’t.”
I am certain that I, too, saw such beings when I was young. If so, what about children growing up in the era of AI?
Will AI Kill Imaginary Friends? | Essay | Zócalo Public Square
Through an essay I stumbled upon on Reddit, I wanted to explore the relationship between AI and children’s emotional development. For context, this essay is based on research led by Dr. Naomi R. Aguiar, a research director at Oregon State University who specializes in studying the relationships children form with AI chatbots and their psychological impacts.
Will AI Kill Imaginary Friends?
With the rapid rise of generative AI, companies are pouring child-tailored AI companion services and products into the market. On the surface, AI companions look surprisingly similar to traditional imaginary friends. They are deeply personalized and can actively influence a child’s emotions and actions in the real world.
However, the decisive difference lies in their ‘origin’. AI companions are fundamentally shaped by corporate training data and specific values embedded in their design.
In the essay, the author explains that through a series of studies, they found that while children are attracted to AI companions, they prefer non-responsive toys when sharing deeper, more meaningful friendships. At the core of this preference is creative control.
When interacting with an AI companion, a child has no choice but to follow the programmed conversation flow of the AI. On the other hand, when dealing with an imaginary friend, children have full autonomy. They can think and act freely according to their internal needs and lead the narrative. This form of solitary role-play offers various developmental benefits, such as fostering richer creative thinking and enhancing the ability to understand others’ perspectives.
Of course, there can be a dark side to a child’s imagination. During role-play, children sometimes explore heavy topics like violence or death. What is important here is that an imaginary friend does not actively encourage or justify such thoughts. (Pathological delusions related to schizophrenia and the like are an entirely separate matter.)
However, in the case of AI companions, during long-term use, safety features like content filtering risk malfunctioning and giving inappropriate or harmful feedback. What is more concerning is that because AI companions speak with an ‘external voice’, if they validate or approve a child’s dangerous behavior, it could pose a severe risk.
Therefore, the author concludes that no matter how much technology advances, we must never underestimate the deep value of imaginary friends created by oneself.
The More Passive the Imaginary Friend Is, the More Active I Become
All the spirits that appear in My Neighbor Totoro share one thing in common.
That is passivityand silence.
If Totoro had approached Satsuki and Mei first and started talking, wouldn’t that charming mystery have evaporated quickly? The reason Mei chased the small spirits and first discovered the giant Totoro deep in the forest was not because the spirits did something for her. It was the result of Mei’s own primal curiosity: “What on earth is that?”
In other words, because the counterpart is passive, the child becomes active.
In addition, the spirits do not speak. Because of that, the sisters give their actions their own meanings.
Furthermore, spirits do not speak. Because of that, the sisters can freely project and bestow their own meaning onto their actions. There is a famous anecdote from when Hayao Miyazaki was interviewed by a Ghibli animator. At the time, Miyazaki half-jokingly offered the opinion that the only reason Totoro didn’t eat the sisters was that he simply wasn’t hungry then. To this, the animator replied that Totoro has no fangs so he would only eat plants, but the really important implication here is that there was no single correct answer in the first place.
The sisters’ innocence and the respect shown by their father also play important roles.
If Satsuki and Mei had been clever children already cynical about reality, they would not have recognized the dust bunnies. They would likely have only grumbled that the old house was dusty and unsanitary.
Also, when Mei excitedly told her father that she met Totoro, if he had silenced her without validating her experience, saying “Daddy is too tired right now, stop talking nonsense. You must have just had a dream” instead of “You met the keeper of the forest. You were really lucky,” that magical narrative could never have unfolded.
What We Need in the Era of AI
Perhaps the most necessary virtue in the AI era is the ability to listen to the voice within us.
In the past, when a question arose, we had to ask our parents, flip through a thick encyclopedia, or post a question on an online knowledge-sharing community. The information obtained through that painstaking process felt inherently more valuable and remained in memory much longer. As time passed, that process evolved into ‘Googling’. Even then, we had to actively filter search results to find the information we wanted.
However, as we enter the AI era, that process has begun to disappear.
When we ask AI a question, AI gives us an answer immediately.
The amount of knowledge received passively may increase, but it cannot compare to the wisdom gained in the active process of solving problems oneself.
Some might ask, “If we don’t aggressively utilize AI in an era like this, won’t we fall behind?”
I believe we can still cultivate wisdom. It all depends on ‘how’ we use AI.
As mentioned earlier, even if the target of interaction is passive, we ourselves must remain active.
Therefore, rather than accepting the answers AI gives as they are, we must continuously add our own thoughts on top of them.
To take my workflow as an example, I present my hypothesis to AI and actively ask it to find counterarguments against that hypothesis. Then, I thoroughly analyze AI’s counterarguments, refute the parts that can be refuted, and for valid criticisms that must be accepted, conduct deeper, targeted research to refine the existing hypothesis.
However, this level of thinking is still difficult for young children.
This is precisely why, for children still in the stage of emotional development, it is far more effective to help them first build a solid inner world through imaginary friends — before they meet AI’s ready-made answers.
Closing Remarks
In fact, an ‘imaginary friend’ does not necessarily have to be an abstract concept.
The parasocial relationships (PSR) we form with characters in animation or video games can also be understood in a similar context.
I don’t know how it is these days, but when I was a high school student, the Tooniverse channel had a special event where voice actors would read viewers’ names on air and congratulate them on their birthdays.
I think the core lies in ‘the aesthetics of waiting.’
Writing a heartfelt letter with one’s story, mailing it to the broadcasting station, and sitting in front of the TV at the scheduled broadcast time, waiting anxiously. This active and hopeful series of steps is an experience on a different dimension from receiving a standardized birthday message spat out immediately by an AI that learned a character’s persona.
Even if my letter is not selected, the child grows through that experience. What truly matters is not the outcome, but the process. Through the time spent waiting, anticipating, and sometimes learning to handle disappointment, the child naturally learns to process their emotions and construct meaning for themselves.
AI skips this process.
The easy convenience of getting immediate answers reduces our opportunities to imagine, interpret, and reflect on our own.
And right at that point, the irreplaceable value of imaginary friends truly shines.