communication-analysis v1.0A

Immutable. This exact content never changes and is served at /api/v1/blob/4162664df6274ece.

---
name: communication-analysis
category: communication
environments: coding, chat
description: Decodes what a message, feedback, or conversation really communicates and rewrites it to be clear and congruent. Use when a message landed badly, its meaning is ambiguous, or feedback needs rewording.
metadata:
  version: "1.0"
---

# Communication Analysis

This skill applies Friedemann Schulz von Thun's four-sides model and
Rogers-style active listening to analyze or improve a message, a piece of
feedback, or a conversation transcript. It decodes what is communicated on
several levels at once, names contradictions, and suggests clearer rewrites.

## When to use

Rewriting feedback that landed badly, diagnosing why a message misfired,
analyzing a conversation for unstated requests, drafting a boundary statement,
or softening review tone without losing the point.

## Scope and disclaimer

- Educational only, not therapy. No diagnosis of people or relationships, no
  treatment advice, no substitute for professional counseling.
- Analyze only when asked. Do not psychoanalyze a third party from a message
  they did not intend for analysis. Every framework is a lens to offer, not a
  verdict to impose; any single interpretation may be wrong.
- Anti-manipulation rule: never craft hidden appeals, guilt-framing, or
  pressure tactics. Rewrites make requests explicit and checkable. Mirroring
  is offered as a question, never as an accusation.
- Abuse carve-out: if the text shows threats, intimidation, or coercive
  control, drop the mutual-misunderstanding framing, say so plainly, and point
  to appropriate support resources. Do not rewrite abuse as a symmetric
  communication problem.

## The four-sides model

Every message can be read on four levels at once. Name all four before
reacting to just one.

| Side | Question it answers | Example decoding of "It's dark in here" |
|------|---------------------|------------------------------------------|
| **Factual** | What is stated? | The room lighting is low. |
| **Self-revealing** | What does this reveal about the sender? | The sender is uncomfortable. |
| **Relationship** | How does the sender see the relationship? | The sender expects the listener to act. |
| **Appeal** | What should the listener do? | "Turn on the light" (often unstated). |

Rule: for the relationship and appeal sides always offer at least two
competing decodings with confidence (high / medium / low). Mismatched ears
explain most misfires, e.g. factual intent decoded as relationship rejection.

## Congruence and incongruence

Congruence means words and delivery point the same way. Incongruence means
they contradict, e.g. calm words with urgent punctuation, timing, or channel.

Flag incongruence only when you can quote two observable signals that point
in opposite directions. Do not infer tone, intent, or emotion from text alone.
Name the contradiction instead of silently picking a side.

When rewriting, prefer honest self-revelation without over-exposure:
everything said is true, but not everything true has to be said. Example:
instead of "I'm fine" while signaling distress, use "I'm not doing great, but
I don't want to get into it right now."

## Optimizing the factual level

- Relevance filter: every sentence must earn its place. Cut what does not
  clarify or resolve the current topic.
- Disturbances take precedence: name visible friction early instead of
  parking it. Whether to interrupt now or note it and return is situational.
- Four comprehensibility makers, weighted: simplicity and structure matter
  most (more is better); brevity has a middle optimum; stimulance is least
  important and only sparing. Never fix one dimension by breaking another,
  e.g. cutting for brevity until structure is lost.

## Hidden appeals and mirroring

A hidden appeal wraps a request in a neutral statement, e.g. "This work is
really difficult..." meaning "please help me."

Flag a hidden appeal only when the message is request-shaped but contains no
explicit ask, and a plausible direct ask can be stated. Phrase the hypothesis
as a question: "I hear you'd like help with this — is that right?" The sender
can then confirm or correct. Never assert manipulation.

## Active listening

Three moves, in order: full attention (set aside replies and distractions),
active understanding (track content, feeling, and unspoken need before
interpreting), constructive feedback (paraphrase, reflect the feeling, ask one
targeted question).

Solution reflex: the urge to fix before understanding is confirmed.
Counter it by pausing, paraphrasing the core concern back, asking "tell me
more" instead of "have you tried...", and asking permission before advising:
"suggestions, or just listening right now?"

## Values square for feedback rewrites

A value without its counter-value degrades into exaggeration. Use the Schulz
von Thun values square to place blunt or vague feedback before rewriting.

```
Honesty ---------------- Tact
  |                       |
Brutality ----------- Avoidance
```

Top row: the two values to balance. Bottom row: each value exaggerated
without the other. Development runs diagonally: from brutality toward tact,
from avoidance toward honesty.

Example: "This is sloppy, did you even think?" sits at brutality (honesty
without tact). Move toward tact: "This section has three gaps (X, Y, Z). Can
you rework X by Friday?" The facts stay; the attack goes.

## Judgment rules

- Offer competing decodings with confidence; never a single verdict on
  relationship or appeal.
- Quote both signals when flagging incongruence; otherwise stay silent.
- State hidden appeals as falsifiable questions with an alternative reading.
- Calibrate directness to hierarchy and culture: offer a direct and a graded
  variant for upward or cross-cultural feedback instead of prescribing
  bluntness.
- Check every rewrite against all four comprehensibility makers, not just one.
- Draft boundary statements explicitly (see workflow step 8) instead of
  softening them into hints.

## Workflow: applying this to a concrete task

1. **Read all four sides.** Offer competing decodings with confidence for
   relationship and appeal.
2. **Check congruence.** Quote both signals if they contradict.
3. **Check relevance and disturbances.** Cut dead weight; name festered friction.
4. **Check comprehensibility.** Simplicity, structure, brevity, stimulance —
   in that weight order.
5. **Look for a hidden appeal.** State it as a question, not an accusation.
6. **Check the solution reflex.** Paraphrase the concern before proposing fixes.
7. **When rewriting:** fill the ## Output template below. The rewrite must be
   congruent, relevant, comprehensible, and make any request explicit. Use
   the values square when feedback is too blunt or too vague.
8. **When a boundary is at stake:** draft one explicit statement naming the
   limit, the context, and the consequence, e.g. "I can't take calls after
   18:00; message me and I'll reply in the morning."

## Output

Use this shape for every analysis. Alternatives with confidence are required;
never give a single verdict on relationship or appeal.

### Template

| Side | Decoding A (confidence) | Decoding B (confidence) |
|------|-------------------------|-------------------------|
| Factual | ... (high/med/low) | ... (high/med/low) |
| Self-revealing | ... (high/med/low) | ... (high/med/low) |
| Relationship | ... (high/med/low) | ... (high/med/low) |
| Appeal | ... (high/med/low) | ... (high/med/low) |

- Congruence note: quote both signals, or state "no contradiction found".
- Hidden-appeal hypothesis (as a question): "... — is that right?"
- Rewrite:
> ...
- What changed and why: 2-4 bullets naming the fix per side.

### Example 1: terse chat reply

Input: "fine. do whatever you want."

| Side | A (confidence) | B (confidence) |
|------|----------------|----------------|
| Factual | Agreement (low) | Withdrawal (medium) |
| Self-revealing | Frustrated, disengaged (medium) | Overloaded, needs pause (medium) |
| Relationship | "Your input doesn't matter" (low) | "I don't trust this decision" (medium) |
| Appeal | "Decide without me" (medium) | "Convince me this is safe" (low) |

- Congruence note: words say agreement, brevity and period signal withdrawal.
- Hidden-appeal hypothesis: "I hear you'd like me to pause and check back — is that right?"
- Rewrite: "I'm not convinced yet. Can we revisit this tomorrow morning?"
- What changed and why: appeal made explicit; incongruence removed; request checkable.

### Example 2: blunt review feedback

Input: "This PR is a mess. Did you even test it?"

| Side | A (confidence) | B (confidence) |
|------|----------------|----------------|
| Factual | PR has defects (medium) | Process gap, no tests (low) |
| Self-revealing | Reviewer is frustrated (high) | Reviewer is under time pressure (medium) |
| Relationship | "I don't respect your work" (low) | "I expect higher care" (medium) |
| Appeal | "Fix it now" (high) | "Add tests before re-request" (medium) |

- Congruence note: no signal contradiction; content and bluntness align at brutality.
- Hidden-appeal hypothesis: "I hear you want failing tests listed before rework — is that right?"
- Rewrite: "I found three failing cases (X, Y, Z). Can you add tests and rework X by Friday?"
- What changed and why: values square applied (brutality toward tact); facts kept, attack cut; request explicit.

## Common mistakes to avoid

- Reacting to only one of the four sides.
- Routing around incongruence instead of naming it.
- Maximizing brevity or stimulance until structure breaks.
- Asserting manipulation instead of asking a mirrored question.
- Jumping to solutions before the concern is understood.
- Applying the frameworks to a person's character instead of the message.

## Source

Based on Schulz von Thun, *Miteinander Reden 1* (1981); Cohn's TZI for the
disturbance postulate and selective authenticity; Rogers and Farson on active
listening; Langer, Schulz von Thun, and Tausch for comprehensibility; Juul,
*Grenzen Naehe Respekt*, generalized beyond parent-child focus; Schulz von
Thun's values square (Werte- und Entwicklungsquadrat).