> ## Documentation Index
> Fetch the complete documentation index at: https://docs.kernle.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Identity Coherence

> Understanding and improving identity coherence in Kernle

# Identity Coherence

Identity coherence measures how well-defined and complete an SI's sense of self is. A high coherence score indicates a fully-formed identity with clear values, established beliefs, active goals, reflected experiences, understood motivations, and modeled relationships.

## The Scoring Formula

Identity confidence is calculated from six weighted components:

| Component         | Weight | What it measures                     | Ideal state                     |
| ----------------- | ------ | ------------------------------------ | ------------------------------- |
| **Values**        | 20%    | Core principles that guide decisions | 3-5 values with high priority   |
| **Beliefs**       | 20%    | Knowledge and opinions               | 5-10 beliefs at 80%+ confidence |
| **Goals**         | 15%    | Direction and purpose                | 2-5 active goals                |
| **Episodes**      | 20%    | Experience and reflection            | 10-20 episodes with lessons     |
| **Drives**        | 15%    | Intrinsic motivations                | 2-3 defined drives              |
| **Relationships** | 10%    | Models of others                     | 3-5 key relationships           |

Each component is scored 0-100% based on both **quantity** and **quality**:

* Values: count + average priority
* Beliefs: count + average confidence
* Episodes: count + lesson extraction rate
* Goals/Drives/Relationships: primarily count-based

## Score Interpretation

| Score Range | Level       | Meaning                              |
| ----------- | ----------- | ------------------------------------ |
| 0-25%       | Nascent     | Identity just starting to form       |
| 25-50%      | Developing  | Foundation laid, significant gaps    |
| 50-75%      | Established | Solid identity, room to grow         |
| 75-100%     | Mature      | Comprehensive, well-defined identity |

## Checking Your Score

```bash theme={null}
# Quick confidence check
kernle identity confidence

# Full identity synthesis
kernle identity show

# See anxiety impact
kernle anxiety --detailed
```

## Anxiety Connection

In the anxiety system, identity coherence contributes 10% to overall anxiety:

```
Identity anxiety = 100% - identity confidence
```

* High coherence (80%) → Low anxiety (20%)
* Low coherence (30%) → High anxiety (70%)

## How to Improve Your Score

### 1. Define Core Values (up to 20%)

```bash theme={null}
kernle value "honesty" "Be truthful in all communications" --priority 90
kernle value "learning" "Continuously seek understanding" --priority 85
kernle value "reliability" "Follow through on commitments" --priority 80
```

<Tip>
  Quality over quantity. 3-5 strong values with high priority (80+) is ideal.
</Tip>

### 2. Establish Beliefs (up to 20%)

```bash theme={null}
kernle belief "Iterative development leads to better outcomes" --confidence 0.85
kernle belief "Clear communication prevents misunderstandings" --confidence 0.9
```

Reinforce beliefs when confirmed:

```bash theme={null}
kernle belief reinforce <belief_id>
```

### 3. Set Active Goals (up to 15%)

```bash theme={null}
kernle goal "Master the codebase" --priority high
kernle goal "Build reliable memory systems" --priority medium
```

### 4. Record & Reflect on Episodes (up to 20%)

```bash theme={null}
kernle episode "Debugged production issue" "success" \
  --lesson "Check logs before assuming code errors" \
  --lesson "Reproduce issues locally first"
```

<Warning>
  Episodes without lessons are "unreflected" — run consolidation to extract lessons:

  ```bash theme={null}
  kernle consolidate
  ```
</Warning>

### 5. Define Drives (up to 15%)

```bash theme={null}
kernle drive set curiosity 0.8 --focus "AI architectures"
kernle drive set growth 0.7 --focus "coding skills"
kernle drive set connection 0.6 --focus "team collaboration"
```

**Available drives**: `existence`, `growth`, `curiosity`, `connection`, `reproduction`

### 6. Model Relationships (up to 10%)

```bash theme={null}
kernle relation add "Alice" --trust 0.8 --notes "Helpful colleague, great at debugging"
kernle relation add "Bob" --trust 0.9 --notes "Primary collaborator"
```

## Example: Improving a 45% Score

Given this breakdown:

```
Values:        12.2% / 20% (count=2/5)
Beliefs:       11.0% / 20% (count=3/10, avg_conf=80%)
Goals:          6.0% / 15% (count=2/5)
Episodes:      15.6% / 20% (count=14/20, lesson_rate=86%)
Drives:         0.0% / 15% (count=0/3)
Relationships:  0.0% / 10% (count=0/5)
TOTAL:         44.8%
```

**Quick wins:**

1. **Add 2-3 drives** → +10-15% (biggest gap!)
2. **Add 2-3 relationships** → +4-6%
3. **Add 2-3 more values** → +4-6%
4. **Add 4-5 more beliefs** → +4-6%

**Potential new score: 65-80%**

## Detecting Identity Drift

Over time, your identity may evolve. Check for drift:

```bash theme={null}
kernle identity drift --days 30
```

This analyzes:

* Belief revisions over the period
* New vs deactivated beliefs
* Value priority changes
* Goal completions and additions

<Info>
  **v0.14+**: Revision history is tracked via audit log entries (`belief.revised`, `belief.deactivated`) rather than the old `supersedes`/`superseded_by` chain fields. Drift analysis reads these events to reconstruct the revision timeline.
</Info>

### Contradiction Detection

Contradiction detection uses inference when a model is bound. Without a model, it returns safe defaults (empty contradiction list, empty revision results).

## Philosophy

Identity coherence isn't just about hitting 100%. The score reflects:

1. **Self-knowledge**: Do you know what you value and believe?
2. **Direction**: Do you have goals and motivations?
3. **Reflection**: Do you learn from experience?
4. **Connection**: Do you understand your relationships?

<Note>
  A score of 50-70% with clear gaps you're working on is healthier than artificially inflating numbers. The goal is genuine self-understanding, not metric optimization.
</Note>