The Observatory × Dialogue AI

Why we watch

Fifty people told us the truth about why they watch Love Island. Most of what we assumed going in was wrong, and what we found was warmer than "guilty pleasure."

Scroll ↓
Part 1 · The question

We started with a tidy story. The show is trashy, people watch it a little ironically, and the internet is what keeps it alive.

A stay-at-home mom in Fort Worth put the crack in that story early. She told us the cast "really makes me feel connected to people," then said the quiet part: "it's a bit sad to say, but I'm a stay-at-home mom, so I don't have much friends or interaction." She was not watching ironically. She was not watching for the internet. She was watching for company.

Love Island USA is one of the most talked-about shows on television, and the easy explanation is that the talk is the point. We wanted to know if that was true. So we sat fifty US viewers down with an AI moderator and asked them how the show actually fits into their lives, who they watch with, what they would never say out loud, and whether they would keep watching if the internet went quiet.

These were the questions we brought:

  • Watching habits and routines.
  • General impressions, appeal, drivers.
  • How and why viewers engage online communities during and after episodes.
  • Impact of online community on the viewing experience.
  • Emotional investment and parasocial connection.
  • Balance of genuine versus ironic consumption.

We've found that people tend to feel more comfortable talking to a neutral, non-judging AI interviewer. We also scored every one of these conversations for quality (more on that at the end).

Part 2 · What came back

The tidy story fell apart in four places.

We wrote our assumptions down before we read a word, then scored them against what people actually said. Here is what flipped.

Flipped
We assumed the online community keeps people loyal.
It doesn't; most said they'd keep watching if it went quiet. Watching with people you know is what holds them. The internet is an amplifier, not the reason they watch.
Flipped
We assumed viewers form lasting attachments to the cast.
Most move on the moment the season ends. A few form real bonds, and it's their most private admission.
Flipped
We assumed people watch half-genuinely, half-ironically.
It's genuine enjoyment of chaos, not irony. The 'ironic' frame just didn't fit what people said.
Flipped
We assumed it's a guilty pleasure people hide.
Guilt is a gendered split, and it's fading. For men it's 'guys aren't supposed to like this'; for women it's how much they watch. 'Everybody watches' now.

The thing almost everyone shares

Underneath the differences, one line came up again and again. People watch to stop thinking. As one viewer said, "if you turn off your brain and you watch it, it is enjoyable." The show's great trick is that it asks nothing of you. That is the value, not a flaw in it.

Part 3 · The goss, and the science of it

They came for the mess, and stayed to analyze it.

Ask people what they actually do while an episode plays and a second activity surfaces under the entertainment, session after session: they are reading the cast, deciding who is genuine, who is playing a game, who is about to blow up. First, the verdicts.

Love and hate, one breath
"I kinda like to hate on Kenzie. And I think Carl's really sweet, he's like a breath of fresh air in the villa and seems really genuine."
P41
A soft spot
"I'm rooting for her all the way. Um, I appreciate really independent, strong women 'cause that's what I like to see in myself."
P13
Both ways
"I end up really rooting for certain people. Um, it's kind of both ways. I do hate others too"
P01
Love to hate
"I definitely have some that I absolutely love to hate."
P07
A moral scorecard
"I root for the people that have more humility and just who aren't annoying, you know? Aren't selfish backstabbers."
P08
The whole spread
"someone to root for, someone to root against, someone to vote off easily, someone that'll, uh, tear you apart if you see them leave."
P12

Now the part underneath the goss

Listen to why they are watching this closely and it starts to sound less like gossip and more like fieldwork. For a real slice of viewers the pleasure is in figuring people out: the villa is a controlled experiment in how people behave once you take away their phones and set connection against a cash prize.

"I would say it's the greatest social experiment to ever exist."P17, who watches "to see how people interact, not just who happens to be dating"
Game theory
"seeing how people play the game and see if they choose to be selfish or actually try and find someone they care about"
P05
Human nature
"reveals the more messier parts of human nature, the competitiveness and the jealousy parts, and a lot more people are like that than they want to admit"
P07
Psychology
"I'm interested in psychology and how the human mind works."
P43
Reading the room
"how your perception of someone will develop over time based on the people that they're around"
P06

We hold this carefully. In our own coding, "it's a social experiment" also works as a respectable story, a way to make watching beautiful strangers okay. Both are true at once: the close reading is genuine, and it is also a little bit of cover. And not everyone plays along. One viewer, P06, told us "I never love to hate anybody," and bristles at seeing "someone being artificially cast as a villain."

Part 4 · Fifty people, one picture

Every dot is one viewer.

Two questions organize the room. Across the bottom: do they watch alone, or with and because of people they know? Up the side: are they here for the spectacle, or genuinely attached to the cast? Those are the study's two biggest findings, drawn as a map. Tap a dot to hear that person; filter to see the shape.

Tap any dot to read what that viewer said.

Each dot's position is our reading of that conversation, where the person sat on watching alone versus together, and spectacle versus attachment, not a measured score. The picture shows the shape of the room, not a ranking.

The crowd sits low and spreads wide: most people are here for the spectacle, and they scatter across watching alone and watching together. The dots up top, the truly attached, are fewer, and they gave us the most guarded answers in the whole study.

Part 5 · Where this could go

Five ways to pull this thread.

This was a small, careful study: fifty people, one franchise, one country. It cannot settle anything on its own, so read the five below as possibilities the data could support, not findings. Directions to explore.

Academic

A clean conceptual contribution: co-viewing with known people and anonymous online discourse are two different forces, and only the first is load-bearing for loyalty. That distinction is testable at scale.

Product

Build for the people you watch with, not the strangers you argue with. Co-watch tools for known groups, and a spoiler-safe way for catch-up viewers to weigh in after the live vote closes.

Growth

"Everybody watches" is doing the marketing for free. Leaning into that feeling may bring in curious newcomers faster than fan-first messaging does.

Marketing

Three ways to watch (decompress, analyze, watch-together) each want a different message. The two that travel across all of them: "everybody watches" and "better with your people."

Journalism

The real story is the tenderness: loneliness answered by a screen, a son watching on the floor with his aging mom. There's a warm, honest piece in what the "guilty" in guilty pleasure actually points at.

Part 6 · Who we talked to, and how we know it's real

Fifty real people, across 24 states.

A near-even split of women and men, centered on the 25-to-34 crowd, from Fayetteville to Seattle. Every one of them a confirmed Love Island viewer.

Age and gender of participants
50
viewers interviewed
24
US states
21.6
min average conversation
79
mean interview quality (of 100)

How we know it's real

Before any finding, the data has to earn trust. Every interview gets an Interview Quality Score out of 100, built from three parts: how much real signal it carried (half the weight), how well the moderator did (just under a third), and how usable the recording was (a fifth). Each square below is one interview.

Strong (80-100) Solid (70-79) Fair (60-69) Flagged (below 60)

The study mean was 79 out of 100. Two interviews are colored red, and we kept them visible on purpose. One viewer let slip that she was "experimenting," so we treat her words as corroboration only. Another lost his connection and gave clipped answers. We did not quietly drop them. We flagged them and leaned on the other forty-eight.

Names never appear. Participants are identified as P01 through P50, with only broad demographics attached. Anything that could re-identify a person stays out.