Skip to content

Index

Citizen Oversight Module

"See power while it's being exercised."

Purpose

Citizen Oversight exists to answer one question:

"Can ordinary people see, understand, and respond to AI power while it is being exercised?"

Not afterward. Not in hindsight. In real time.

What Citizen Oversight IS

  • A visibility layer for AI system behavior
  • A comprehension system that translates signals into plain language
  • A participation framework for civic engagement
  • A memory architecture that prevents institutional amnesia

What Citizen Oversight IS NOT

  • ❌ Not surveillance of individuals
  • ❌ Not censorship of content
  • ❌ Not centralized control
  • ❌ Not punishment infrastructure

Mobius watches systems with power, not people.

Oversight Pillars

1. Visibility (See)

Citizens can: - View integrity metrics (MII) - See intent declarations - Track system behavior over time - Observe changes, not just snapshots

No raw surveillance. Only behavioral transparency.

2. Comprehension (Understand)

Every signal must be: - Human-readable - Plain-language summarized - Contextualized historically

If citizens can't understand it, oversight has failed.

3. Participation (Act)

Citizens can: - Flag integrity concerns - Submit evidence - Request review escalation - Participate in governance rituals - Earn MIC for civic oversight work

Participation is opt-in, accountable, and recorded.

4. Memory (Remember)

Oversight records: - Cannot be deleted - Cannot be quietly altered - Cannot be "PR-washed" away

This prevents rebranding abuse and institutional amnesia.

Documentation Index

Document Description
DASHBOARD.md Dashboard wireframes and layout
MICROCOPY.md UI text, tooltips, and warnings
API_CONTRACTS.md Public API specifications
TRUST_SCORES.md Citizen trust and anti-brigading
MIC_INCENTIVES.md Reward mechanics for oversight

Citizen Roles

Role Access Capabilities
Observer Read View dashboards, subscribe to updates
Auditor Read + Flag Flag anomalies, attach evidence
Steward Full Governance participation, review escalations

All actions are logged. No anonymous power.

What Citizens Oversee

CAN oversee: - Integrity drift - Unexplained behavior changes - Threshold violations - Repeated harm patterns - Accountability gaps

CANNOT: - Spy on individuals - Access private data - Control AI outputs directly - Censor speech

Example Oversight Flow

1. MII drops below baseline
2. ZEUS flags anomaly
3. Public dashboard updates
4. Citizens see:
   - what changed
   - when
   - why it matters
5. Auditors submit evidence
6. Steward review triggered
7. Outcome logged permanently

No secrecy. No panic. No erasure.

The Civic Compact

Mobius does not promise safety. It promises visibility.

And visibility is what gives people the chance to act before the ending is written.

Design Philosophy

Non-Negotiables

Before any dashboard design:

  1. No raw personal data — Never exposed
  2. No model internals — Black box preserved
  3. No outputs surveillance — Content not monitored
  4. No individual tracking — People are not watched

This dashboard shows system behavior over time, not people.

If a citizen can't explain what they're seeing in one sentence, the UI has failed.

"Mobius doesn't decide what's right. It makes wrongdoing impossible to hide."

We heal as we walk. — Mobius Substrate