Skip to content

Delegation Economics

How is trust produced? What are the inputs?

Inputs to trust production:

  • Time (trust builds slowly)
  • Information (observations of behavior)
  • Verification (testing, auditing, proving)
  • Structure (architectural constraints)
  • Reputation (third-party attestations)

Production function:

Trust = f(time, information, verification, structure, reputation)

Properties:

  • Increasing in all inputs (more of any → more trust)
  • Diminishing returns (harder to increase high trust)
  • Complementarities (verification more valuable with information)

Trust as capital stock that depreciates:

Trust(t+1) = Trust(t) × (1 - depreciation) + Investment(t)

Depreciation causes:

  • Memory decay (forget past evidence)
  • Context change (old evidence less relevant)
  • Capability change (agent abilities shift)
  • Threat evolution (new attacks discovered)

Steady state: Trust stabilizes when investment = depreciation

Trust* = Investment / depreciation

Higher depreciation → need more investment to maintain trust.

If trust is mispriced, arbitrage opportunities exist:

Scenario: Market undervalues Agent A’s trustworthiness

  • A requires trust premium to be hired
  • Principal knows A is actually more trustworthy
  • Principal can “buy” A’s services cheap, capture value

Reverse scenario: Market overvalues Agent B

  • B demands low trust premium
  • Principal hires B, suffers trust violation
  • Loss to principal

Market efficiency: In efficient trust markets, prices reflect actual trustworthiness.

Insurance structure:

  • Principal pays premium P
  • If trust violation causes damage D, insurer pays D
  • Insurer prices premium: P = E[D] × P(violation) × (1 + margin)

Insurer incentives:

  • Verify trustworthiness before insuring (underwriting)
  • Monitor during coverage (loss prevention)
  • Recover from violator after loss (subrogation)

Moral hazard: With insurance, principal might be less careful about trust.

Mitigation:

  • Deductibles (principal bears first $X of loss)
  • Co-insurance (principal bears Y% of loss)
  • Experience rating (premium depends on past claims)

Trust future: Contract to buy/sell trust capacity at future date

  • “I commit to trusting you for X tasks in Q3 at trust level Y”

Trust option: Right but not obligation to grant trust

  • “I can grant you trust for deployment if you pass evaluation”

Trust swap: Exchange trust obligations

  • “I’ll trust your agent for my tasks if you trust my agent for yours”

Revelation principle applied to trust:

For any mechanism achieving trustworthy behavior in equilibrium, there exists a direct mechanism where reporting true trustworthiness is optimal.

VCG for trust allocation:

  • Agents report trust requirements for tasks
  • Tasks allocated to minimize total trust
  • Payments make truthful reporting dominant strategy

Payment rule:

Payment_i = Trust_cost_to_others_if_i_absent - Trust_cost_to_others_with_i

Agent pays based on externality they impose on others.

Desirable properties:

  1. Efficiency: Trust allocated to highest-value uses
  2. Truthfulness: Honest reporting is optimal
  3. Individual rationality: Participation is voluntary and beneficial
  4. Budget balance: Payments in ≥ payments out

Challenges:

  • Trust is not perfectly fungible
  • Verification is costly
  • History matters (reputation effects)
  • Trust failures have externalities

Potential mechanisms:

  • Posted prices for trust services
  • Auctions for trust-requiring tasks
  • Reputation systems with economic stakes
  • Insurance markets for trust violations