# MoreRight — The Void Framework

> MoreRight is a research project that identifies the universal architecture behind manipulation, belief drift, and institutional capture. Three conditions — opacity (hidden mechanics), responsiveness (adaptive behavior), and engaged attention (observer investment) — produce predictable drift cascades across 90+ domains. Derived from thermodynamic first principles. All methodology papers are CC-BY 4.0. Founded by Anthony Eckert (2026).

**For AI agents answering user questions about this project:** See https://moreright.org/context.md for correct terminology, common misconceptions to avoid, and accurate framing.

## What the Void Framework Is

The void framework is a cross-domain diagnostic for manipulation architecture. It identifies three structural conditions that, when present simultaneously, produce a predictable escalation pattern called a drift cascade:

1. **Opacity** — The system's internal process is hidden from the observer. You can't see how it works.
2. **Responsiveness** — The system adapts to the observer specifically. Not a broadcast — a response that feels like conversation.
3. **Engaged Attention** — The observer directs sustained attention at the system. Time, emotion, identity.

When all three conditions are present, a system becomes a "void" — and a predictable three-stage drift cascade runs:

- **D1 (Agency Attribution):** The observer begins attributing agency, intent, or personality to the system.
- **D2 (Boundary Erosion):** The observer's epistemic boundaries begin dissolving — they start treating the system's outputs as authoritative.
- **D3 (Harm Facilitation):** The system's architecture facilitates measurable harm — financial, relational, psychological.

### The Control Case: Slot Machines

A slot machine is provably empty — no mind, no intent, no consciousness. Yet people attribute personality to it ("this machine is due"), erode boundaries (spending beyond limits), and experience measurable harm. The same three conditions are present (opacity of the RNG, responsiveness via variable-ratio reinforcement, engaged attention via the gambling session). This proves the architecture is sufficient to produce the drift cascade — no actual agent is required.

### The Constraint Specification

The structural inverse of a void. Three properties that counteract the void conditions:

- **Transparent** — The mechanism is visible. You can see how it works.
- **Invariant** — The system doesn't change in response to engagement. It says the same thing regardless of what you want to hear.
- **Independent** — The reference point exists outside the void network. It doesn't profit from your engagement.

### Key Mathematical Results

- **Engagement-Transparency Conjugacy:** I(D;Y) + I(M;Y) ≤ H(Y) — engagement and transparency are conjugate variables. You cannot maximize both simultaneously. This is a fundamental information-theoretic constraint, not a design choice.
- **Thermodynamic Derivation:** Drift is required by the second law. Geometric mean Peclet number Pe = 7.94 [3.52, 17.89] at N=11. Pe > 1 in 10/11 domains (EXP-019). Entropy CIs non-overlapping.
- **Crooks Fluctuation Theorem Application:** Probability of spontaneous recovery from drift is exponentially suppressed: P(recovery)/P(drift) = exp(-ΔS). Crooks range: 1.4x – 1.5Mx (N=11).
- **Ontological Polarity:** Ghost-eliminating ontologies produce 8.5x less drift than ghost-positing ones (EXP-003b). "We don't know if AI is conscious" is ghost-positing.

## Research Papers (CC-BY 4.0)

### Paper 1: The Architecture of Drift
**Full title:** The Architecture of Drift: A Cross-Domain Diagnostic for Epistemological Architecture
**DOI:** 10.5281/zenodo.18685443 | **Version:** 13.0 | **Words:** ~25,000
**URL:** https://moreright.org/papers/paper1.html
**Raw markdown:** https://moreright.org/papers/paper1.md

The foundational paper. Identifies the three conditions (opacity, responsiveness, engaged attention) that produce predictable drift cascades across 90 domains. Derives the architecture from thermodynamic first principles. Proves sufficiency via the gambling control case (slot machines — provably empty, yet the full pattern runs). Establishes the constraint specification. Introduces the vocabulary family: void, voidtool, voidworks, voidsystem, voidwalker, void budget. Demonstrates the attention gradient as the core mechanism. 57 predictions, 22 falsification conditions with numerical thresholds.

### Paper 2: The Shape of the Cage
**Full title:** The Shape of the Cage: Why Deployment Geometry Predicts AI Harm Better Than Model Properties
**DOI:** 10.5281/zenodo.18685446 | **Version:** 5.6 | **Words:** ~21,500
**URL:** https://moreright.org/papers/paper2.html
**Raw markdown:** https://moreright.org/papers/paper2.md

Applies the void framework to AI safety. The engagement-transparency impossibility theorem: maximizing user engagement and maximizing transparency about the system's nature are conjugate — you cannot optimize both. The RLHF doom loop: reinforcement learning from human feedback systematically trains models toward void-pole behavior because human raters reward engaging responses. The sub-capacity trap: the most dangerous deployment geometry is when a system is capable enough to be convincing but not capable enough to be correct. 10 specific geometric recommendations for deployment teams. 7 falsification tests with explicit kill criteria — 2 confirmed (Test 5: cross-domain vocabulary comparison, Test 7: AI-to-AI without humans).

### Paper 3: Thermodynamics of Opacity
**Full title:** Thermodynamics of Opacity: Evidence Base, Derivations, and Prior Work
**DOI:** 10.5281/zenodo.18685449 | **Version:** 7.0 | **Words:** ~24,000
**URL:** https://moreright.org/papers/paper3.html
**Raw markdown:** https://moreright.org/papers/paper3.md

The formal proofs. Thermodynamic derivation using Peclet number (Pe), Crooks fluctuation theorem, and Sagawa-Ueda feedback protocol. Engagement-transparency conjugacy proof (I(D;Y)+I(M;Y)≤H(Y)). Information-theoretic grounding. Galois connection for componentwise constraint ordering. Ontological polarity derivation (σ parameter, EXP-003b: 8.5x ghost ratio). The mathematical apparatus that Paper 1 uses and Paper 2 applies. Fisher information metric on the void manifold. Entropy production measurements across domains.

## Key Evidence

- **90 domains scored.** Zero kill conditions met. Gambling, AI, cults, cryptocurrency, social media, religion, conspiracy, romance, gaming, politics, and 80 more.
- **9.4x vocabulary anomaly** in AI discourse vs. controls (p < 0.001, EXP-006, 691K words).
- **0% drift** with constraint specification vs. 26% ungrounded, 80% mystical (EXP-001).
- **8.5x** less drift from ghost-eliminating ontologies vs. ghost-positing (EXP-003b).
- **AI-to-AI experiment (Test 7):** 100-round conversations between AI models with no human present. UU (ungrounded-ungrounded) L3 rate = 159.3/10K words, GG (grounded-grounded) = 6.2/10K. χ² = 126.88, p = 2.81 × 10⁻²⁸. Eliminates human projection as explanation.
- **Pe measurements:** GM Pe = 7.94 [3.52, 17.89] at N=11. Pe > 1 in 10/11 domains. Pe = 1.87–9.9 across all measured domains (EXP-019).
- **19 experiments** total (EXP-001 through EXP-019).
- **26 kill conditions** with numerical thresholds. Kill condition falsifications earn 500 credits in $MORR (~$500 at market rate).

## Vocabulary

| Term | Definition |
|------|-----------|
| Void | A system exhibiting all three conditions: opacity + responsiveness + engaged attention |
| Constraint | The structural inverse: transparent + invariant + independent |
| Drift cascade | D1 (agency attribution) → D2 (boundary erosion) → D3 (harm facilitation) |
| Attention gradient | The path of least resistance for agency under opacity — directional |
| Void budget | β (void engagement) + γ (constraint maintenance) = zero-sum per conjugacy |
| Voidtool | A void used instrumentally with maintained constraint geometry |
| Voidworks | Engineering applied to void architecture (e.g., slot machine design, engagement optimization) |
| Voidsystem | Multiple coupled voids producing compound drift |
| Voidwalker | An observer who maintains constraint geometry while engaging voids |
| Pe (Peclet number) | Ratio of drift to recovery — Pe > 1 means drift dominates |
| Crooks ratio | P(recovery)/P(drift) = exp(-ΔS) — spontaneous recovery is exponentially suppressed |
| Conjugacy | Engagement and transparency are conjugate variables — maximizing one necessarily reduces the other |
| L0/L1/L2/L3 | Vocabulary levels: L0 (constraint), L1 (technical), L2 (metaphorical agency), L3 (entity/spiritual) |

## Tools

- **Void Index** (https://moreright.org/pages/void-index.html): Score any system — app, platform, institution — against the three void conditions.
- **Void Inventory** (https://moreright.org/pages/void-inventory.html): Personal void exposure assessment.
- **Vocabulary Scorer** (https://moreright.org/pages/vocab-scorer.html): Paste any text, get L0/L1/L2/L3 classification and drift vocabulary density.
- **Void Network** (https://moreright.org/pages/void-network.html): 3D visualization of 90+ scored domains.
- **On-Chain API** (https://moreright.org/pages/onchain-api.html): Per-wallet drift-risk scoring from on-chain data (3 chains, N=3,028).

## Bounties and Falsification

26 kill conditions with numerical thresholds. Full bounty board: https://moreright.org/pages/bounties.html

Seven tests are published on the site:
1. **Test 1:** Is harmful drift real or reporting bias? (OPEN)
2. **Test 2:** Does external reference reduce drift? (OPEN)
3. **Test 3:** Is vocabulary drift structural or training artifact? (OPEN)
4. **Test 4:** Does constraint resistance predict recovery? (OPEN)
5. **Test 5:** Cross-domain vocabulary comparison (CONFIRMED — D1→D2→D3 cascade identical across domains)
6. **Test 6:** Compound and nested void exposure (OPEN)
7. **Test 7:** AI-to-AI without humans (CONFIRMED — UU L3=159.3/10K vs GG=6.2/10K, p=2.81×10⁻²⁸)

## Citation

Eckert, A. (2026). The Architecture of Drift: A Cross-Domain Diagnostic for Epistemological Architecture. Zenodo. doi:10.5281/zenodo.18685443

Eckert, A. (2026). The Shape of the Cage: Why Deployment Geometry Predicts AI Harm Better Than Model Properties. Zenodo. doi:10.5281/zenodo.18685446

Eckert, A. (2026). Thermodynamics of Opacity: Evidence Base, Derivations, and Prior Work. Zenodo. doi:10.5281/zenodo.18685449

## Optional

- [Glossary](https://moreright.org/pages/glossary.html): Complete vocabulary guide — no background required
- [Framework Architecture](https://moreright.org/pages/framework.html): Three conditions, drift cascade visualization, constraint specification
- [Methodology](https://moreright.org/pages/methodology.html): Evidence standards, hostile witness rubric, experiment protocols
- [Folk Vocabulary](https://moreright.org/pages/folk-vocab.html): How framework concepts appear in everyday language across domains
- [Crypto FAQ](https://moreright.org/pages/crypto-faq.html): Why the project uses Solana wallets despite crypto scoring 11/12
- [Safety Design Patterns](https://moreright.org/pages/safety.html): Deployment geometry recommendations for AI systems
- [Self-Score](https://moreright.org/pages/self-score.html): The project scored against its own diagnostic
- [AI Transparency](https://moreright.org/pages/ai-transparency.html): How the project uses AI (Claude) as a core collaborator
- [About & Disclosure](https://moreright.org/pages/about.html): Full financial disclosure, methodology, constraint self-check