Skip to content

Epistemic & Coordination Solutions

The same AI capabilities that create epistemic risks can potentially be harnessed for defense. AI can:

  • Verify claims faster than humans can
  • Aggregate distributed knowledge effectively
  • Coordinate large groups around shared goals
  • Detect manipulation and deception
  • Preserve and organize knowledge

This section catalogs concrete strategies and technologies that could help societies maintain epistemic integrity and coordinate effectively in an AI-saturated environment.


Using AI to verify claims, detect manipulation, and authenticate content:

  • AI-Assisted Fact-Checking — AI systems that help verify claims at scale
  • Content Authentication — Provenance, watermarking, and verification systems
  • Deepfake Detection — AI systems to detect synthetic media

Harnessing distributed human knowledge and judgment with AI augmentation:

  • Prediction Markets — Markets for aggregating probabilistic beliefs
  • AI-Augmented Forecasting — Combining AI and human judgment for better predictions
  • Deliberation Platforms — AI-assisted democratic deliberation
  • Wisdom of Crowds — Mechanisms for aggregating distributed knowledge

Tools for enabling large-scale cooperation and coordination:

  • Commitment Mechanisms — Credible commitment devices for AI governance
  • Multi-Stakeholder Platforms — Coordination across governments, labs, and civil society
  • AI-Assisted Negotiation — AI systems that help parties find agreements

Foundational systems for maintaining shared knowledge:

  • Knowledge Graphs — Structured knowledge representation with provenance
  • Epistemic Auditing — Tools for evaluating information quality
  • AI-Human Hybrid Systems — Designs that combine AI and human strengths

AI ThreatPotential AI Defense
AI-generated misinformationAI detection, provenance verification
Personalized manipulationAI that explains manipulation attempts
Scale of fake contentAI verification at matching scale
Coordination attacksAI-assisted collective defense
Expertise erosionAI-augmented human expertise

The offense-defense asymmetry (generation easier than verification) is real, but:

  • Verification compounds: Once verified, information stays verified
  • Trust is valuable: Verified sources become focal points
  • Defense can coordinate: Attackers often can’t coordinate as well
  • AI advantages scale: Defensive AI improves with resources

Neither humans nor AI alone are sufficient:

  • AI: Speed, scale, consistency
  • Humans: Judgment, values, accountability
  • Together: Combine strengths, check weaknesses

Systems must work despite:

  • Active attackers trying to subvert them
  • Edge cases designed to fool AI
  • Coordination among bad actors
  • Evolving attack techniques

Systems work when:

  • Participants benefit from honest behavior
  • Gaming is harder than sincere participation
  • Costs fall on bad actors, not good ones
  • Long-term reputation matters

Start small and build:

  • Begin with low-stakes applications
  • Demonstrate reliability before scaling
  • Build track record over time
  • Allow for correction and improvement

Avoid single points of failure:

  • Multiple independent verifiers
  • Diverse approaches and methods
  • Open protocols others can implement
  • No single entity controls the system

CategoryExamplesMaturity
Prediction marketsPolymarket, Metaculus, ManifoldGrowing
Fact-checking AIClaimBuster, Full Fact AIEarly
Content authenticationC2PA, Content CredentialsEmerging
Deliberation platformsPolis, All Our IdeasNiche use
Knowledge platformsWikipedia, Semantic ScholarEstablished
GapImportance
Cross-platform verificationHigh
Real-time deepfake detectionHigh
AI-assisted diplomatic negotiationMedium
Global epistemic infrastructureHigh
Incentive-aligned knowledge curationHigh

  • Can AI-enabled defense keep pace with AI-enabled offense?
  • What institutional structures are needed to deploy these solutions?
  • How do we bootstrap trust in new verification systems?
  • Can coordination solutions work across geopolitical divides?
  • What research is most neglected relative to importance?