Parasocial imbalance: who you watch vs. who watches back
Your archive records the direction of attention you gave and the direction you received - why one-sided digital watching feels like a relationship, and how measured asymmetry looks without any feelings attached.
Parasocial relationships - one-sided bonds formed by watching someone who never
watched you back - feel real because attention itself is rewarding, not because
it is reciprocal. Your export can measure the asymmetry directly: stories you
viewed against story interactions you received, follows against followers,
sent messages against received ones. The app reports the two directions side by
side and stops there; nobody's feelings are computed from a ratio.
You know the feeling: you have watched someone's stories for months, maybe
years, and could not name one thing they know about you. In 1956 Horton and
Wohl called the first half of that a parasocial relationship - the audience
member's bond with a performer who does not know they exist.1 Social
media made ordinary people into performers for people they never met, and made
watching into a daily, low-cost, private act.
What social media did not change is the arithmetic. Attention still only
accumulates on one side until something reciprocates it. The difference is that
an export finally lets you see both numbers.
The asymmetry your archive can actually measure
Directional evidence in a typical export includes:
- Stories you viewed vs. story interactions you received. The archive
carries your watch history as its own activity kind ("Stories you viewed")
separate from story likes and sticker reactions you gave, while received
reactions land on the other side of the ledger.2 - Who you watch vs. who watches you. The audience data you get is bounded
by what the platform hands you; what you gave is recorded on your side
regardless. - Following vs. followers. Fans - accounts that follow you but you do not
follow back - are computed as a set difference, and so is the mirror image,
unfollowed-but-stayed.3 - Sent vs. received messages. Per-conversation balance, reply-rate and
median reply time are all derived from sender tags on message rows.
Each pair is the same shape: attention flowing one way, measured independently
of attention flowing the other. Nothing about the pairs is inherently sad -
an asymmetric pair can be a creator you admire, a public account, or a friend
whose notifications you muted years ago. The asymmetry is a fact about
direction, and direction is not a verdict.
Why one-sided watching registers as a relationship
Three findings from the research literature explain why watching feels like
bonding even when nothing comes back:
- Parasocial bonds use the same social machinery as real ones. Neuroimaging
work on media personas found that fans' brains respond to their favourite
celebrity with some of the same signals triggered by close others - the
bond is processed socially, not as object-attachment.4 - Mere exposure plus familiarity does quiet work. Repeatedly seeing a
face, a name, a story frame builds familiarity, and familiarity reads as
closeness. The feed delivers repetition automatically; you supply the
attribution. - The interaction is designed to feel interactive. Replies, reaction
stickers and "close friends" labels make an asymmetric channel feel like a
conversation. The reciprocity is staged - which is exactly what the word
parasocial was coined to describe.1
None of this is a personal failing. It is a predictable outcome of one-direction
broadcast channels wrapped in two-direction interface language.
What the product does with the two directions
The accounts directory and per-account pages show both sides per person
without collapsing them: your interactions with them, their follows of you,
your searches for them, and the tier computed from real exchanges - core,
active, at-risk, casual, dormant - where at-risk is a measurable drop in their
replies, not an emotional prediction.5 The people-interactions surfaces
aggregate the directional totals: what you gave, what you received, where the
ledger leans.
What it deliberately does not do:
- No "who's obsessed with you" leaderboard and no "you're obsessed with them"
shame list. Directional counts exist; social verdicts do not.
- No parasocial score. Nothing in the codebase classifies a relationship as
one-sided-and-unhealthy, because that judgment requires context the file
cannot hold - a one-sided watch of a sibling's stories is still family. - No inference about anyone's awareness or intentions. The archive cannot know
whether they saw your follow, your reply or your story reply, and the views
never imply that it can.
Reading your own asymmetry honestly
If you want to look, do it in this order:
- Check the sample first. Watch history and story audiences are subject to
the platform's export windows - counts are floors, not totals.
- Separate categories. A story you viewed, a like you sent and a message
you sent are three different intensities of attention; the taxonomy keeps
them apart precisely so you don't sum them into one number that means
nothing. - Look at direction before volume. "I gave 300 units, they gave 20" says
less than "I gave every week for three years; they gave twice, then
stopped" - and the timeline views, not the totals, carry that second story. - Resist the urge to send a test message. The classic unhealthy response
to seeing asymmetry is to manufacture reciprocity. The archive is a reading
instrument; it does not require any response from you or them.
Can this app tell me who is watching my stories?
Only to the extent your export contains it - the platform gives you your own
side of story activity, and audience lists arrive only where the platform
provides them (for example, for stories still inside their window). What you
viewed is recorded richly; who viewed you is bounded by the export, and the app
shows the difference rather than smoothing it over.
Does a one-sided ratio mean the relationship is fake?
No. It means attention flowed one way in the recorded channels. Public figures,
quiet friends, lurkers who care and never reply - all produce the same ratio.
The number establishes direction, not meaning; the meaning needs your context.
Why can't you calculate a 'parasocial index' from my archive?
Because the inputs would be counts pretending to be psychology. Any index would
encode somebody's guess about when one-sided attention becomes unhealthy - a
guess that depends on your life, your history and the other person's, none of
which the file contains. The product measures direction and balance; it refuses
to name what they feel like.
I watched someone's stories for years and they never followed me back. Is that in here?
The direction is: your watch records are on your side, their follow status is
in the connections snapshot. Both appear on the same per-account page so you
can see the pair. What is not in the archive - whether they ever noticed,
whether they muted you, whether they meant anything by it - and the app will
not invent it.
Related reading
The people-and-interactions inventory for every directional record the export
contains; the relationships post for how tiers and decay are computed from real
exchanges; the shadow-diary post for the other side of the equation - attention
you paid where nobody, including the watched person, could see it.
1: Horton, D. and Wohl, R. - "Mass Communication and Para-Social Interaction,"
Psychiatry (1956) - the original formulation of the one-sided performer/
audience bond; cited as research framing.
4: Published fMRI studies of celebrity fandom report activation of
reward and social-cognition networks when fans view preferred celebrities -
cited here as background on why watching registers socially, not as a claim
this product verifies.
2: packages/shared/src/analytics/activity-taxonomy.ts - storyViewed
("Stories you viewed", group consumed) and storyLike ("Story likes &
sticker reactions", group reacted) are separate kinds; activity-taxonomy.ts
also maps received story interactions from storyInteractions.
3: packages/shared/src/analytics/archive-compare.ts:172 - fans are
computed as followers.filter(u => !following.has(u)); the same file computesunfollowedButStayed as the mirror set difference.
5: packages/shared/src/analytics/engagement-intel.ts:175 - tier
thresholds (core >= 60, active >= 35, casual >= 15, else dormant) and
the at-risk rule at engagement-intel.ts:173 (score, recent-vs-prior reply
rate, days since last contact) - all computed from message rows.
Footnotes
- horton
- taxonomy
- fans
- neuro
- intel