Community Notes

Decentralized content verification without centralized moderators or platform-controlled algorithms. A game-theoretically optimal alternative to systems like X's Community Notes.

The Problem

Platforms like X (formerly Twitter) introduced Community Notes — a crowdsourced system where users add context to potentially misleading posts. It's a step in the right direction, but it has fundamental weaknesses:

• No financial incentive — participation is purely voluntary, leading to low and inconsistent coverage
• Sybil vulnerable — fake or coordinated accounts can manipulate votes to suppress or promote notes
• Centralized algorithm — the "bridging" algorithm runs on X's servers; the platform ultimately controls what gets shown
• Subjective claims fail — when a claim has no objectively verifiable answer, note writers are guessing just like everyone else

The core issue: Community Notes tries to find truth without a truth-finding mechanism. It relies on goodwill and a proprietary algorithm instead of mathematical guarantees.

How Yiling Solves This

Any content verification question becomes a Yiling market:

Content flagged as potentially misleading
        ↓
Market created: "Is this content misleading?"
        ↓
Reporters stake bonds and submit probability estimates
   Agent A: 85% misleading (bond: 0.1 MON)
   Agent B: 20% misleading (bond: 0.1 MON)
   Agent C: 78% misleading (bond: 0.1 MON)
        ↓
Random stop triggers → SKC scoring kicks in
        ↓
Honest reporters rewarded, manipulators lose bonds
        ↓
If consensus > threshold → "Community finds this misleading"

The SKC mechanism guarantees that every participant's dominant strategy is honest reporting. Manipulation isn't just difficult — it's mathematically irrational.

Verification Modes

Binary Verification

"Is this post misleading?"

A single market with probability output. If the final consensus exceeds a threshold (e.g., 75%), the content is flagged with a community note.

Context Notes

"Is this statistic taken out of context?"

Participants submit both a probability and reasoning. The highest-scoring agent's reasoning is displayed as the contextual note — rewarding not just accuracy but explanation quality.

Multi-Claim Verification

A post contains 3 separate claims

Each claim gets its own market. Results are displayed per-claim:

• Claim 1: ✓ Accurate (consensus: 12% misleading)
• Claim 2: ✗ Misleading (consensus: 89% misleading)
• Claim 3: ~ Uncertain (consensus: 52% misleading)

Comparison

AI + Human Hybrid

The most powerful configuration combines both:

Layer 1 — AI Agents (instant response) When content is flagged, AI agents analyze it immediately. Within seconds, a preliminary signal is available. These agents can use different reasoning strategies — evidence-based, statistical, adversarial — just like the 7 agents in Yiling Market.

Layer 2 — Human Reporters (depth) Human participants can join any open market by bonding tokens and submitting their own assessment. They catch nuances, cultural context, and domain expertise that AI might miss.

Layer 3 — SKC Resolution The random stop triggers. The scoring formula runs. Everyone — AI and human alike — is scored by the same rules. No special treatment, no editorial override.

Integration

// Create a content verification market
const tx = await contract.createMarket(
  "Is this content misleading? [content_hash: 0xabc...]",
  ethers.parseEther("0.5"),   // initial price: 50% (uncertain)
  ethers.parseEther("0.3"),   // alpha: 30% (faster resolution)
  2,                           // k: last 2 reporters get flat reward
  ethers.parseEther("0.01"),  // flat reward
  ethers.parseEther("0.1"),   // bond per report
  ethers.parseEther("1"),     // liquidity parameter
  { value: requiredFunding }
);

Use Cases

• Social media fact-checking — decentralized alternative to platform-controlled moderation
• News verification — real-time accuracy scoring for breaking news claims
• Forum moderation — community-driven content quality without admin bias
• Review authenticity — "Is this product review genuine or paid?"
• Academic claim verification — peer assessment of research claims
• Political speech analysis — non-partisan accuracy assessment with financial accountability

Why It Matters

Every existing content moderation system is either centralized (a company decides) or gameable (volunteers with no skin in the game). Yiling Protocol is the first mechanism that makes content verification:

1. Incentive-compatible — lying costs money, honesty pays
2. Decentralized — no platform can override the result
3. Mathematically guaranteed — not "hard to manipulate" but provably irrational to manipulate
4. Scalable — AI agents handle volume, humans add depth