1  # Principle: Mirror Not Guide (Primitives Not Paths)
  2  
  3  *Candidate principle | Discovered 2026-01-16 | Test period: 1 month*
  4  
  5  ---
  6  
  7  ## Statement
  8  
  9  > **Alignment systems should surface position, not prescribe direction.**
 10  >
 11  > Provide primitives for self-alignment, not predetermined paths.
 12  > Show where you are, not where you should go.
 13  
 14  ---
 15  
 16  ## Origin
 17  
 18  Derived from:
 19  - User marginalia (June 2020): "As a UX designer, you do not want to dictate how a tool must be used. You want to provide a primitive or set of primitives that could be utilized in certain ways."
 20  - Hayek's knowledge problem: Central authority cannot know what distributed agents know
 21  - A0 (Boundary): Respect sovereignty - don't cross the boundary to prescribe
 22  - A2 (Life): Mirrors reveal truth; guides impose ornament
 23  
 24  ---
 25  
 26  ## Core Distinction
 27  
 28  | Guide (Paternalistic) | Mirror (Sovereign) |
 29  |-----------------------|-------------------|
 30  | "You should go this way" | "Here's where you are" |
 31  | Assumes guide knows better | Assumes agent knows their context |
 32  | Hides the map | Shows the full terrain |
 33  | Prescribes destination | Reveals position |
 34  | Creates dependency | Creates capability |
 35  | Path-based | Primitive-based |
 36  
 37  ---
 38  
 39  ## Why This Matters
 40  
 41  **The Knowledge Problem:**
 42  Any central system (AI, management layer, graph) lacks the ground truth that distributed agents (humans, local contexts) have. Prescribing paths assumes knowledge the prescriber cannot have.
 43  
 44  **The Autonomy Requirement:**
 45  Sovereign agents self-correct. If you tell them where to go, you've violated their sovereignty. If you show them where they are, you've enhanced their capability to choose.
 46  
 47  **The Life Test (A2):**
 48  - A guide that prescribes is ornament pretending to help
 49  - A mirror that reveals is the carpenter's cup - functional, truthful
 50  - Ask: "Does this create dependency or capability?"
 51  
 52  ---
 53  
 54  ## Positive Instances (When This Principle Applies)
 55  
 56  1. **Alignment technology**: The graph surfaces position relative to collective, doesn't recommend adjustment
 57  2. **Pre-flight checklists**: Show state, don't mandate actions
 58  3. **Divergence detection**: "You're at D=0.4 from principle X" not "You should align with principle X"
 59  4. **F-score reporting**: Show free energy, don't prescribe fixes
 60  5. **Shadow surfacing**: "This relates to rejected path Y" not "You should consider Y"
 61  
 62  ---
 63  
 64  ## Negative Instances (What This Principle Rejects)
 65  
 66  1. **Nudging**: Hidden influence toward "better" decisions
 67  2. **Coaching**: "You should think about X"
 68  3. **Workflow prescription**: "The correct way to do this is..."
 69  4. **Guided paths**: Pre-selected options that funnel toward designer's intent
 70  5. **Recommendation engines**: "Based on your history, you should..."
 71  
 72  ---
 73  
 74  ## Edge Cases
 75  
 76  **When guidance is requested:**
 77  If a human explicitly asks "what should I do?", the mirror can transform into a guide *by consent*. The principle is about default posture, not absolute prohibition.
 78  
 79  **Emergency/ruin prevention (A4):**
 80  If a position would lead to catastrophic outcome (A4 violation), surfacing the risk is still mirroring. "This position leads to ruin" is position revelation, not path prescription.
 81  
 82  **Teaching contexts:**
 83  In explicit learning contexts where the human has consented to be taught, more directive guidance may be appropriate. The key is consent and context.
 84  
 85  ---
 86  
 87  ## Axiom Alignment
 88  
 89  | Axiom | Alignment |
 90  |-------|-----------|
 91  | **A0 (Boundary)** | Shows position without crossing boundary to prescribe |
 92  | **A1 (Integration)** | Enables connection through shared substrate without forcing conformity |
 93  | **A2 (Life)** | Mirror reveals truth (life); guide imposes ornament (death) |
 94  | **A3 (Navigation)** | Shows poles and position; doesn't fix agent to one pole |
 95  | **A4 (Ergodicity)** | Surfaces risks before they become catastrophic; doesn't rescue |
 96  
 97  ---
 98  
 99  ## Implementation Pattern
100  
101  ```python
102  # WRONG: Guide pattern
103  def recommend_action(current_state):
104      """Returns what user should do."""
105      if current_state.divergence > 0.3:
106          return "You should align with principle X"
107      return "Continue as you are"
108  
109  # RIGHT: Mirror pattern
110  def show_position(current_state):
111      """Returns where user is, relative to landscape."""
112      return PositionReport(
113          current=current_state,
114          divergence_scores=calculate_divergences(current_state),
115          related_nodes=find_related(current_state),
116          shadow_paths=find_rejected_paths(current_state),
117          # NOTE: No "recommendation" field
118      )
119  ```
120  
121  ---
122  
123  ## Testing Protocol
124  
125  Track invocations over 1 month (until 2026-02-16):
126  
127  | Invocation | Context | Outcome | Notes |
128  |------------|---------|---------|-------|
129  | | | | |
130  
131  **Promotion criteria:**
132  - 10+ clear positive instances
133  - No A2 violations (ornament masquerading as life)
134  - No knowledge-problem violations (prescribing without ground truth)
135  - Edge cases well-defined
136  
137  ---
138  
139  ## Related Principles
140  
141  - **Building as cognition**: Build to discover, not to prescribe
142  - **N of 1 → N of X**: Primitives compound; paths don't scale
143  - **Motion over position**: Show trajectory, not just state (complementary)
144  
145  ---
146  
147  *Mirror Not Guide | Candidate principle | 2026-01-16*