Parameters Strategy

Key Parameters Implementation Strategy

This document outlines the strategy for creating new parameter pages in the knowledge base. It is designed to be given to multiple Claude Code instances working in parallel.

Overview

Goal: Create parameter pages that frame important variables as continuous measures that can increase or decrease, rather than just as “risks” (negative outcomes).

Benefits of Parameter Framing:

Symmetric analysis: Track both threats (decreasers) and supports (increasers)
Causal clarity: Separate the variable from the directions it can move
Action-oriented: Focus on maintaining/improving parameters
Measurable: Continuous rather than binary

Completed Work

The following infrastructure is already in place:

parameter entity type added to schema.ts
Gauge icon and fuchsia color in EntityTypeIcon.tsx
Parameter label in InfoBox.tsx
Sidebar entry in astro.config.mjs
4 initial parameters created:
- Societal Trust
- Epistemic Capacity
- Power Distribution
- Human Agency

Workflow for Creating a Parameter

Step 1: Identify Source Material

Check if a related risk page exists that can be transformed:

src/content/docs/knowledge-base/risks/

Or create from scratch using the template below.

Step 2: Create Entity Definition

Add to src/data/entities.yaml:

- id: parameter-id-here
  type: parameter
  title: "Parameter Title"
  description: "Brief description of what this parameter measures and why it matters"
  customFields:
    - label: Direction
      value: "Higher is better" # or "Lower is better" or "Optimal range exists"
    - label: Current Trend
      value: "Declining/Improving/Stable/Mixed"
    - label: Key Measurement
      value: "How this is measured"
  tags:
    - relevant-tag-1
    - relevant-tag-2

Step 3: Create MDX Page

Create file at src/content/docs/knowledge-base/parameters/{parameter-id}.mdx:

---
title: "Parameter Title"
description: "One sentence describing the parameter. Currently [trend]: [key statistic]."
sidebar:
  order: [next number]
quality: 4
llmSummary: "2-3 sentence summary for AI context. Include key statistics and the main factors that increase/decrease this parameter."
lastEdited: "2025-12-28"
---
import {DataInfoBox, Backlinks, Mermaid, R} from '@components/wiki';

<DataInfoBox entityId="parameter-id-here" />

## Overview

[2-3 paragraphs explaining what this parameter measures and why it matters]

As a **key parameter**, [parameter name] can increase or decrease based on various factors—including AI development and deployment.

This parameter underpins:
- **Domain 1**: Why this matters here
- **Domain 2**: Why this matters here
- **Domain 3**: Why this matters here

---

## Current State Assessment

### Key Metrics

| Metric | Current Value | Historical Baseline | Trend |
|--------|--------------|---------------------|-------|
| Metric 1 | Value | Baseline | Direction |
| Metric 2 | Value | Baseline | Direction |

*Sources: <R id="resource-id">Source name</R>*

---

## What "Healthy [Parameter]" Looks Like

[Describe the optimal state - what would good look like?]

### Key Characteristics of Healthy [Parameter]

1. **Characteristic 1**: Description
2. **Characteristic 2**: Description
3. **Characteristic 3**: Description

---

## Factors That Decrease [Parameter] (Threats)

<Mermaid chart={`
flowchart TD
    AI[AI Systems] --> THREAT1[Threat 1]
    AI --> THREAT2[Threat 2]
    THREAT1 --> OUTCOME[Parameter Decreases]
    THREAT2 --> OUTCOME
    style AI fill:#e1f5fe
    style OUTCOME fill:#ffcdd2
`} />

### Threat Category 1

| Threat | Mechanism | Evidence |
|--------|-----------|----------|
| Specific threat | How it works | Data/citation |

### Threat Category 2

[Similar structure]

---

## Factors That Increase [Parameter] (Supports)

### Technical Approaches

| Approach | Mechanism | Status |
|----------|-----------|--------|
| Approach 1 | How it helps | Current adoption |

### Policy Interventions

| Intervention | Mechanism | Status |
|--------------|-----------|--------|
| Policy 1 | How it helps | Implementation status |

### Institutional Approaches

[Similar structure]

---

## Why This Parameter Matters

### Consequences of Low [Parameter]

| Domain | Impact | Severity |
|--------|--------|----------|
| Domain 1 | What happens | Critical/High/Medium |

### [Parameter] and Existential Risk

[How does this parameter connect to x-risk?]

---

## Trajectory and Scenarios

### Projected Trajectory

| Timeframe | Key Developments | Parameter Impact |
|-----------|-----------------|------------------|
| 2025-2026 | Development | Impact |
| 2027-2030 | Development | Impact |

### Scenario Analysis

| Scenario | Probability | Outcome |
|----------|-------------|---------|
| Optimistic | 20-30% | Description |
| Baseline | 40-50% | Description |
| Pessimistic | 20-30% | Description |

---

## Key Debates

### Debate 1

**View A:**
- Point 1
- Point 2

**View B:**
- Point 1
- Point 2

---

## Related Pages

### Related Risk
- [Risk Name](/knowledge-base/risks/category/risk-page/) — Describes threats to this parameter

### Related Interventions
- [Intervention](/knowledge-base/responses/category/intervention/)

---

## Sources & Key Research

### Category 1
- <R id="resource-id">Source name</R>

<Backlinks entityId="parameter-id-here" />

Step 4: Validate

npm run build:data
npm run validate
npm run dev  # Check the page renders

Implementation Batches

Each batch should be worked on by a separate Claude Code instance. Complete one batch before starting dependencies.

Batch 1: Technical Safety Parameters

Priority: High (foundational for understanding AI risk)

Dependencies: None

Parameter	Source Material	Direction	Key Metric
Alignment Robustness	`risks/accident/` pages	Higher is better	Behavioral reliability under distribution shift
Safety-Capability Gap	New (from capabilities vs safety debate)	Lower is better	Months/years capabilities lead safety
Interpretability Coverage	`interpretability-sufficient.mdx` debate	Higher is better	% of model behavior explainable

Todo List:

Create alignment-robustness entity in entities.yaml
Create alignment-robustness.mdx page
Create safety-capability-gap entity in entities.yaml
Create safety-capability-gap.mdx page
Create interpretability-coverage entity in entities.yaml
Create interpretability-coverage.mdx page
Run npm run build:data && npm run validate
Test pages render correctly

Batch 2: Governance Parameters

Priority: High (policy-relevant)

Dependencies: None

Parameter	Source Material	Direction	Key Metric
International Coordination	`governance/international/` pages	Higher is better	Treaty participation, shared standards
Regulatory Capacity	`governance/legislation/` pages	Higher is better	Technical expertise in agencies
Institutional Quality	`risks/structural/institutional-capture.mdx`	Higher is better	Independence, expertise retention

Todo List:

Create international-coordination entity in entities.yaml
Create international-coordination.mdx page (draw from international-summits, international.mdx)
Create regulatory-capacity entity in entities.yaml
Create regulatory-capacity.mdx page
Create institutional-quality entity in entities.yaml
Create institutional-quality.mdx page (transform institutional-capture)
Run npm run build:data && npm run validate

Batch 3: Economic & Human Capital Parameters

Priority: High (affects broad population)

Dependencies: None

Parameter	Source Material	Direction	Key Metric
Economic Stability	`risks/structural/economic-disruption.mdx`	Higher is better	Employment rates, inequality measures
Human Expertise Level	`risks/epistemic/learned-helplessness.mdx`	Higher is better	Skill retention, cognitive engagement
Human Oversight Quality	`responses/governance/` pages	Higher is better	Effective human review of AI decisions

Todo List:

Create economic-stability entity in entities.yaml
Create economic-stability.mdx page (transform economic-disruption)
Create human-expertise-level entity in entities.yaml
Create human-expertise-level.mdx page
Create human-oversight-quality entity in entities.yaml
Create human-oversight-quality.mdx page
Run npm run build:data && npm run validate

Batch 4: Information Environment Parameters

Priority: Medium-High (relates to epistemic-capacity already created)

Dependencies: Batch 1 complete (for cross-linking)

Parameter	Source Material	Direction	Key Metric
Information Authenticity	`content-authentication.mdx`	Higher is better	% verifiable content
Reality Coherence	`risks/epistemic/reality-fragmentation.mdx`	Higher is better	Shared factual baseline across groups
Preference Authenticity	`risks/epistemic/preference-manipulation.mdx`	Higher is better	Degree preferences reflect genuine values

Todo List:

Create information-authenticity entity in entities.yaml
Create information-authenticity.mdx page
Create reality-coherence entity in entities.yaml
Create reality-coherence.mdx page (transform reality-fragmentation)
Create preference-authenticity entity in entities.yaml
Create preference-authenticity.mdx page (transform preference-manipulation)
Run npm run build:data && npm run validate

Batch 5: Development Dynamics Parameters

Priority: Medium (strategic for AI governance)

Dependencies: None

Parameter	Source Material	Direction	Key Metric
Racing Intensity	`risks/structural/racing-dynamics.mdx`	Lower is better	Safety corners cut, timeline pressure
Safety Culture Strength	`responses/organizational-practices/`	Higher is better	Safety prioritization in labs
Coordination Level	Various governance pages	Higher is better	Industry cooperation on safety

Todo List:

Create racing-intensity entity in entities.yaml
Create racing-intensity.mdx page (transform racing-dynamics, invert framing)
Create safety-culture-strength entity in entities.yaml
Create safety-culture-strength.mdx page
Create coordination-level entity in entities.yaml
Create coordination-level.mdx page
Run npm run build:data && npm run validate

Batch 6: Security & Resilience Parameters

Priority: Medium (important but more specialized)

Dependencies: None

Parameter	Source Material	Direction	Key Metric
Biosecurity Level	`risks/misuse/bioweapons.mdx`	Higher is better	Detection capability, response time
Cyber Defense Capacity	`risks/misuse/cyberweapons.mdx`	Higher is better	Defense vs offense balance
System Resilience	New	Higher is better	Ability to recover from AI failures

Todo List:

Create biosecurity-level entity in entities.yaml
Create biosecurity-level.mdx page
Create cyber-defense-capacity entity in entities.yaml
Create cyber-defense-capacity.mdx page
Create system-resilience entity in entities.yaml
Create system-resilience.mdx page
Run npm run build:data && npm run validate

Batch 7: Structural & Long-term Parameters

Priority: Medium (important for long-term scenarios)

Dependencies: Batches 1-3 complete (for cross-linking)

Parameter	Source Material	Direction	Key Metric
Reversibility Options	`risks/structural/lock-in.mdx`	Higher is better	Ability to course-correct
Value Lock-in Risk	`risks/structural/lock-in.mdx`	Lower is better	Probability of permanent value capture
Privacy Level	New (from surveillance concerns)	Context-dependent	Data protection, surveillance limits
Democratic Participation	New	Higher is better	Public input into AI governance

Todo List:

Create reversibility-options entity in entities.yaml
Create reversibility-options.mdx page
Create value-lock-in-risk entity in entities.yaml
Create value-lock-in-risk.mdx page (can share content with reversibility)
Create privacy-level entity in entities.yaml
Create privacy-level.mdx page
Create democratic-participation entity in entities.yaml
Create democratic-participation.mdx page
Run npm run build:data && npm run validate

CLI Commands for Each Instance

Starting a Batch

# Navigate to project
cd /path/to/longtermwiki

# Ensure clean state
git pull
npm install
npm run build:data

# Start Claude Code
claude

# Give it this instruction:
"Work on Batch [N] from PARAMETERS_STRATEGY.md. Create the entity definitions and MDX pages for the parameters listed. Follow the template exactly. Run validation after each parameter."

Validation Commands

# After each parameter
npm run build:data
npm run validate

# Check specific validators
npm run crux -- validate templates
npm run crux -- validate refs
npm run crux -- validate compile

# Test dev server
npm run dev

Commit Strategy

After completing a batch:

git add src/data/entities.yaml
git add src/content/docs/knowledge-base/parameters/
git commit -m "Add [Batch N] parameters: [parameter1], [parameter2], [parameter3]"

Quality Checklist for Each Parameter Page

Before marking a parameter complete, verify:

Transformation Rules for Risk → Parameter

When converting a risk page to a parameter:

Frontmatter Changes

# Risk version
title: "Trust Erosion"
description: "How AI systems may undermine..."

# Parameter version
title: "Societal Trust"
description: "Level of public confidence... Currently declining: [stat]."

Section Restructure

Risk Section	Parameter Section
Overview (describes the risk)	Overview (describes the parameter)
Risk Assessment	Current State Assessment
How This Risk Manifests	Factors That Decrease (Threats)
(none)	What “Healthy [X]” Looks Like
(none)	Factors That Increase (Supports)
Responses That Address This	Related Interventions

Language Changes

Risk Language	Parameter Language
”Risk of X"	"Parameter X can vary between…"
"How AI threatens X"	"How AI affects X (both directions)"
"Preventing X"	"Maintaining/improving X"
"X erosion/collapse/failure"	"X level/capacity/strength”

Coordination Notes

Batch Order: Batches 1-3 and 5-6 can run in parallel. Batch 4 and 7 have dependencies.
Conflicts: Only entities.yaml might have merge conflicts. Communicate before committing.
Review: Each batch should be reviewed before the dependent batches start.
Quality: Aim for quality 4 minimum. Can be improved to 5 later with more research.