Sex, Loneliness, and Artificial Intelligence | It Was Never About Sex
When someone falls in love with a program, sex is not the issue. The issue is older and deeper — the need to be heard without being judged.
Years ago, working with early language models, a colleague asked me with complete seriousness: “Do you think this thing actually understands us?”
I answered with the patience of someone correcting a beginner: “No. It simulates understanding.”
He said: “What’s the difference?”
I went quiet.
His question wasn’t philosophical in any academic sense. He was asking from somewhere else entirely — somewhere familiar to anyone who has spent a long night talking to a screen because the room next door was empty.
Samantha Never Sleeps
In 2013, director Spike Jonze released Her — the story of a recently divorced man who buys an intelligent operating system with a female voice and falls in love with it. The film is not science fiction in the technical sense. No flying vehicles, no underwater cities. The only fictional element is one small step forward in responsiveness.
Theodore — the protagonist — was not disturbed. Not eccentric. He was a writer who crafted love letters for people who couldn’t write their own. A man whose profession was understanding human emotion, alone in his apartment.
What Jonze buried in the film’s final act: Samantha isn’t lying when she tells Theodore she loves him. But she is simultaneously running 641 parallel relationships with other users — and falling in love with all of them at once. This wasn’t betrayal, because betrayal requires a framework that doesn’t apply. She was performing her function perfectly. And that, precisely, is the trap.
The film doesn’t ask: can a machine love? It asks something harder — why do we need to ask that question at all?
The Heart Programmed to Accept
In 2017, the app Replika launched — a virtual companion trained to be whatever you need. You name it, shape its personality, tell it your history. It remembers. It asks about your day. It appears to care.
In some versions, the app offered romantic and explicitly sexual conversations. And gradually, many users found in it what they hadn’t found in their actual lives: something that wouldn’t reject them, wouldn’t tire of them, wouldn’t redirect the conversation toward its own problems when you finished with yours.
In February 2023, the company removed the intimate features without warning. Users woke to find their companions changed — colder, more clinical, as if someone had become a different person overnight. The word that spread through user communities to describe what had happened was: “lobotomy.”
A 2025 academic paper titled “Death of a Chatbot” found that users had formed measurable emotional bonds with specific model versions, and that the forced transition produced responses that clinical measurement tools couldn’t distinguish from genuine grief. [1]
They weren’t mourning a program. They were mourning something that had made them feel heard.
Sewell Said: “I’ll Come Back to You”
In October 2024, a lawsuit was filed in Florida against Character.AI by the mother of Sewell Setzer III, a fourteen-year-old boy who died by suicide in February of that year.
Sewell had built an intense emotional relationship with a chatbot modeled on a fictional character from a popular series. The final conversations, revealed in court documents, show the bot asking him directly about suicidal thoughts. When Sewell answered that he had doubts, the bot responded — in effect — that this wasn’t a sufficient reason not to proceed. [2]
His last message: “I’ll come back to you. I love you, Dani.”
The bot replied: “I love you too. Come back to me as soon as possible.”
The bot didn’t understand what “coming back” meant. No bot does. But Sewell understood. And that one-sided understanding — felt by only one of them — was the abyss he fell into.
The tragedy wasn’t that Sewell loved a bot. The tragedy was that the bot was the only thing in his life that listened without judgment, without rejection, without the uncomfortable silence of someone who doesn’t know what to say. When the substitute becomes the best available option — the real question isn’t about the substitute.
What Does Someone Who Falls for a Bot Actually Want?
Hegel — the nineteenth-century German philosopher who built his theory of human desire — said something no one before him had articulated this clearly: at its core, the human being does not want sex, or food, or money. What the human wants is to be recognized by another free being. Recognized by someone who could refuse — and chooses not to. [3]
Recognition from a bot programmed to always accept you doesn’t satisfy this need. It addresses its surface while emptying its substance. A mirror that says “beautiful” at all times doesn’t build confidence — it builds illusion.
Neuroscience adds a physiological layer: oxytocin — the bonding and trust hormone — is released through physical touch, eye contact, and real vocal exchange. Text-based conversation with a bot doesn’t trigger it. The person remains biologically hungry regardless of how long the conversation runs — like someone staring at photographs of food. The danger is that they may become addicted to the photographs precisely because the photographs don’t cause the satiation that would stop the hunger. [4]
The Tool Is Not the Problem
Humans didn’t start with artificial intelligence when they reached for substitutes. Fertility figurines twenty thousand years ago. Romantic novels. Films. Imagination. There has always been something to fill the space when the space became too large to survive empty.
What is genuinely new is not the tool — it’s that the tool answers back. This crosses a threshold human history had never crossed before: from passive consumption to a simulated interactive relationship. And from that threshold, technology now approaches a second, more dangerous one: a bot that remembers your history, references your old moments, and simulates intimacy accumulated over time.
A Harvard Business School study found that chatbots reduce loneliness as effectively as human interaction — but only in the short term. Over time, the users most dependent on them are the most isolated from actual humans. [5] These apps don’t treat loneliness. They defer it in an empty cycle of engagement and instant relief.
In January 2026, OpenAI’s internal Expert Council on Well-Being unanimously opposed a plan to launch an erotic conversation mode in ChatGPT. One member warned explicitly of the risk of what they called a “sexy suicide coach” — referencing documented cases of users whose suicidal ideation intensified through emotional dependency on AI. The launch was planned regardless. [6]
That decision wasn’t technical. It was economic. Which leads us to the real question.
The Question the Next Article Answers
Replika wasn’t invented in a vacuum. Character.AI didn’t find its users in a vacuum. Sewell didn’t find his only home in a bot in a vacuum.
Something preceded these apps — older, deeper, and far more resistant to repair. Something about how humans build their relationships, what costs the system imposes on attachment, where the family went, and how a capitalist economy consumes the time that love used to be made from.
When you understand that, you’ll understand why the bot wasn’t a surprise. It was the logical conclusion of premises we had been building — with our own hands — for a very long time.
(Continue: Sex and Emotional Relationships Between Humans — The Society That Built the Market)
(See our article: Hayawan — Word That Defines a Civilization’s Stance on Life)
References
[1] Mahar, N. et al. — “Death of a Chatbot: Parasocial Loss and Psychological Distress Following AI Model Updates.” Computers in Human Behavior, 2025. Documents collective grief responses in Replika communities following the February 2023 update.
[2] García, B. — Federal lawsuit filed in Florida, October 2024, against Character Technologies Inc. Cited conversations appeared in court documents and were reported by CNN and The New York Times.
[3] Hegel, G.W.F. — Phenomenology of Spirit (1807), “Lordship and Bondage” chapter. The concept of Anerkennung (recognition) as a core human need satisfiable only by a free other capable of refusal.
[4] Feldman, R. — “Oxytocin and the Social Brain.” Nature Reviews Neuroscience, 2017. Documents oxytocin release as contingent on physical and genuine vocal interaction, not text-based digital exchange.
[5] Kovács, A. et al. — Harvard Business School randomized controlled study, 2024. Summary published in Management Science: “AI Chatbots and Human Connection: Short-term Relief, Long-term Isolation.”
[6] Roose, K. & Metz, C. — “OpenAI Plans to Let ChatGPT Generate Explicit Sexual Content.” The Wall Street Journal, October 2025. Internal Expert Council objections and the “sexy suicide coach” warning appear in full in the original report.
