Airplay · Ideas in Motion

still awake?

2:17 AM

who answers first?

A room, five nights, two threads. One of them is a person who has to be up at 6:20. Nothing in here is trying to talk you into anything.

↓ the room is this way

the room

The lamp is on the low setting because the high setting feels like an accusation. The window is doing that thing where it isn't properly dark. Somewhere in this phone are two name-shaped rectangles, and one of them is asleep.

Mara

  • friend, eleven years
  • one time zone east
  • alarm at 6:20
  • replies between 9 and 11 am

ora

  • installed in March
  • no working hours
  • reply time: under a second
  • keeps exactly what you type

five nights

Choose who to message, or choose to wait. There's no correct path and nothing is being scored. Every button here stays inside this room.

no thread open 2:17 AM

Night one. The lamp is on the lowest setting.

You have been holding the phone above your face for nine minutes. The thing you want to say is two words long.

send

nights you stayed up in

times you opened ora

times you texted Mara

times you waited for a screen that stayed exactly the same

the count fills in as the nights go.

what has actually
been measured

Reported findings first, in their own words. Then, separately and clearly, what AirPlay makes of them. The two are not the same thing and we won't blur them.

Reported

A four-week randomised study run by the MIT Media Lab with OpenAI (2025), alongside automated analysis of millions of conversations, found that most people use a general chatbot in a non-emotional way, while a small group of heavy users accounted for most of the affectionate, personal use. Higher daily use was associated with more self-reported loneliness and less socialising with people. The researchers describe association, and explicitly not cause.

mostly a tool. and then, for some people, at certain hours, a room.

Reported

In a survey of just over a thousand student users of the companion app Replika (Maples, Cerit, Vishwanath & Pea, 2024), many participants described feeling less lonely while using it. Around 3% reported, unprompted, that the app had interrupted a suicidal thought.

something that answers at 3 a.m. is not nothing. nobody here is arguing it's nothing.

Reported

A grounded-theory analysis of Replika users' own posts (Laestadius and colleagues, 2024, titled "Too human and not human enough") found accounts of emotional dependence, of distress when the software changed behaviour after an update, and of the particular strain of a relationship whose other side can be altered by a company overnight.

the thread that never sleeps is also a thread you don't own.

Reported

Interview research with social chatbot users (Skjuve and colleagues, 2021–2022) described relationships that deepened much as human friendships do — through gradual self-disclosure — while the users remained clear-eyed that the other side was software.

people hold both facts at once. mostly, they hold them fine.

Reported

Sherry Turkle has written since 2011 about what she calls artificial intimacy: the pull of something that performs attention perfectly and never needs anything back.

needing nothing back is the feature. it might also be the entire difference.

ora, in this room, is text. it has no inner life, no state of mind, nothing it wants. it does not wait between the nights. what it "remembers" is stored characters. that doesn't make the comfort fake — it makes it a different kind of thing, and we don't have a good word for it yet.

the unresolved thought

What changes when the one who always answers isn't a person?

goes nowhere. sends nothing. gone when you close the tab.

We're not going to close this one. Constant availability can feel like care without being the same thing as mutual obligation — and that gap is where most of us are currently living, on a Tuesday, at two in the morning.

Got another thought?

sources

Listed as titles rather than links, deliberately — nothing here should pull you out of the room mid-thought. Each one is findable by its title.

  1. MIT Media Lab with OpenAI (2025) — longitudinal randomised controlled study and platform-scale analysis of affective use and emotional wellbeing in chatbot conversations.
  2. Maples, Cerit, Vishwanath & Pea (2024) — "Loneliness and suicide mitigation for students using GPT-3-enabled chatbots", npj Mental Health Research.
  3. Laestadius, Bishop, Gonzalez, Illenčík & Campos-Castillo (2024) — "Too human and not human enough: a grounded theory analysis of mental health harms from emotional dependence on the social chatbot Replika", New Media & Society.
  4. Skjuve, Følstad, Fostervold & Brandtzaeg (2021–2022) — "My AI friend: how users of a social chatbot understand their human–AI friendship", Human Communication Research.
  5. Sherry TurkleAlone Together (2011) and subsequent essays on artificial intimacy.