The perception you don't get to see from inside the conversation.
The core Chat X-Ray tells you what happened in the chat. This addon tells you how it landed — on four scientific instruments: Kenny's Social Relations Model for meta-perception, the Johari Window for what's open, blind, hidden, and unknown, Fiske's SCM impression profile, and Leary's Sociometer for relational-value signals. Every score is anchored in specific messages from the chat — no surveys, no guesses.
The perception you don't get to see from inside
You over-estimate what you show them on three axes and under-estimate what you show them on two. The result is that they read you as more contained and less invested than you actually feel — and you calibrate every next message to that misread. This is the loop 'How They See Me' isolates.
Reflected Appraisal Accuracy: 58 / 100
Three overlaid views per axis
Blue = your view of you. Gold = their view of you (behavioural inference). Red = your guess at their view. Where blue and gold diverge = self-other agreement gap. Where red and gold diverge = meta-error.
Meta-error vs agreement gap per axis
Meta-error (primary) = how wrong your guess about their view is. Agreement gap (gold) = how far apart you and they actually are. Big meta-error + small gap = miscalibrated worry. Small meta-error + big gap = accurately seen misalignment.
You think your warmth is fully visible. Their behaviour reads you as slightly reserved — the drafted register masks the warmth in your intent.
The largest agreement gap. You feel deeply invested; they see effort but not the full weight because you compress it into logistics.
Your retract-and-resend hides the vulnerability you feel you're offering. They receive the final version, not the drafted one.
Biggest meta-error. You think your frustration is barely visible; theirs reads you as visibly frustrated more often than you realise (July cluster).
The most accurate meta-perception. Care lands because it shows up in tone regardless of drafting.
You think you look composed; the frequency of check-ins reads to them as need, even when each individual message reads calm.
Reflected Appraisal Accuracy is 58/100 — moderate. You are calibrated well on Care and Warmth, and poorly on Frustration and Neediness (the two axes you most try to control). The pattern is classic: the more you edit an emotion before sending, the worse your read of how it lands.
What's open, blind, hidden, unknown
You show up on time and follow through. They read this, you know it.
Apr 3 — "you're the one person I don't have to double-check on plans"
The private-jokes vocabulary is co-authored and mutually acknowledged.
May 18 — "our stupid little code words are half the reason i miss you"
You ask about their work, family, projects — this is visible and they name it back.
Jun 2 — "you're the only one who actually asks about the small stuff"
Drafted messages arrive polished; the felt intensity behind them doesn't cross the wire.
Mar 22 — Visible edit down to "anyway. how are you." — they receive calm; you sent controlled.
You believe you kept July civil. They mapped July as a hard month before you did.
Jul 27 — "is this the thing we do now" — you experienced this as one hard message; they experienced it as the fourth in a pattern.
Each individual ping reads warm; the rhythm reads anxious.
Jun–Jul — 4.3 initiations/week from you vs 1.2 from them — cumulative asymmetry.
You still track it monthly. They treated it as closed by May.
Jun 12 — "still think about April sometimes" — the only callback on your side, sent alone.
Latency, initiation ratio, callback rate — you track these implicitly. They have no idea.
Aug 14 — "is it weird that we can be this warm in a single exchange and then not talk properly for a week" — one-time surfacing of the constant internal ledger.
Your language contains three concrete future scenarios; you have never shared any of them.
May 30 — "if we can't fix this weekend we can't fix anything" — the only near-visible reference; retracted by follow-up message.
The chat cannot distinguish between 'they can't do closeness right now' and 'they don't want this level of closeness'. This is the single unresolved question the report can't answer without them naming it.
You have never held the line long enough to see what they do in the void. The information doesn't exist yet.
Neither party has named what a successful repair would feel like. Both are guessing.
How you land on six impression axes
Six-axis impression radar (0–10)
Warmth × Competence is the Fiske SCM backbone; Agency, Reliability, Desirability, and Depth complete the relational-value picture.
The impression profile leads with Competence + Reliability (both >8) and lags on Depth and Desirability. Translation: you land as the trusted, capable partner they can rely on — and slightly less as the person they cannot imagine losing. That is the single most actionable perception gap in the report.
High, but muted by drafted register. Warmth in intent > warmth as received.
Read as reliable, capable, articulate — the single strongest impression axis.
You initiate, plan, and steer. Reads as agency; sometimes reads as pressure.
The highest axis. You are the person who follows through. This is the impression pillar.
In-moment high, sustained-moderate. Desire language on their side is warm but not urgent.
Perceived as thoughtful but slightly opaque — the Hidden quadrant is what caps this.
Aggregate index +2.1 / ±5
Six signal channels (−5 rejecting … +5 valuing)
Each channel is a distinct way a person signals relational value. The shape between channels is the story — steady inclusion with falling prioritisation is a very different picture from uniformly warm signal.
Apr 3 — "you're the one person I don't have to double-check on plans" — inclusion signal, unchanged across corpus.
H1 vs H2 — Priority-language ("come first", "before anything else") occurs 7× in H1 and 1× in H2.
Feb 14 — "my person" — used 3× in H1, 0× in H2.
Jun–Aug — Their word-count per exchange drops 41% between Q2 and Q3.
Jul 29 — "That wasn't fair of me and I know it landed badly. I'm sorry." — repair capacity intact even as prioritisation slides.
corpus — Turn-toward rate 54% (below the 65–70% stable benchmark).
The sociometer index sits at +2.1 (mild positive) but the shape matters more than the number. Inclusion and Repair are intact and stable — you are inside their circle and they can come back after harm. Prioritisation, Investment, and Bid-response are all trending down — you are inside the circle but sliding toward its edge. This is the mechanism behind the felt sense of 'we're fine but something is off'.
How stable are these perception estimates?
| Estimate | Point | 95% CI | Stability |
|---|---|---|---|
| Perception gap (overall) | 4.1 | 3.6 – 4.6 | 87% |
| Reflected Appraisal Accuracy | 58 | 52 – 64 | 83% |
| Sociometer index | 2.1 | 1.6 – 2.5 | 86% |
| Meta-error: Frustration | 2.2 | 1.7 – 2.7 | 81% |
| Impression: Desirability | 6.4 | 5.9 – 6.9 | 84% |
Add "How They See Me" to any Chat X-Ray for $9.99
Available as an addon after your core report finishes. Same 24h retention, same evidence-first analysis, same scientific instruments — applied to your actual conversation.