The Observatory × Dialogue AI · Data Snapshot

50 Love Island viewers, by the numbers

A descriptive landscape of who these viewers are and how they vary, the quantitative companion to the ethnographic read. Where the qualitative work argues, this section maps.

50
AI-moderated interviews (moderator "Mia")
21.6 min
mean interview length (14.3-31.0)
24
US states represented
79.1/100
mean Interview Quality Score

Who we talked to

A near-even gender split and a center of gravity in the 25-34 bracket. Every participant is a confirmed Love Island viewer; the screener required it.

Age and gender distribution
Interview duration histogram
Top states
NY
6
CA
6
TX
5
FL
4
PA
4
MD
3
NC
2
TN
2

What they watch

Reality TV is universal, but these are omnivores, documentaries and comedies nearly rival reality, and two-thirds watch sports while over half follow news. They are not a niche.

Genres watched
Reality shows watched in past 6 months
Love Island versions watched
Most watch beyond the US edition, the UK and All Stars seasons have real reach in this sample.

How they watch, the two social layers

This is the chart that reframes the study. "By myself" leads, but co-viewing with known people (friends, a partner, family) stacks up right behind it. That known-person layer is what the qualitative work finds load-bearing for loyalty. The anonymous online layer sits separately, as an amplifier.

Watch settings
Season 8 engagement level
Engagement splits cleanly: 21 are appointment-viewers who vote live, another 21 catch up on their own schedule.

Sizing the qualitative, with caution

Qualitative prevalence undercounts: not mentioning a theme is not evidence against it. These are therefore floor counts ("coded in at least N of 50"), not point estimates, useful for relative weight, not precision.

Coded prevalence floor counts
The single largest pattern is chaos and spectacle over rooting for romance, followed by brain-off decompression as the core value. Genuine parasocial attachment is a real but minority pattern, and, per the ethnographic read, the most hidden one.

Data quality

A healthy dataset: mean IQS 79.1/100, with idiosyncratic personal detail throughout (a viewer on the couch "with my cats," another entering the call while breastfeeding, a man who watches "on the floor" with his aging mother). Two interviews are flagged, one authenticity tell, one tech breakdown, and handled with caution downstream.

IQS distribution