The Agency Protocol uses specialized evidence types to support different mechanisms of trust. Evidence can support promises, verify assessments, or establish probabilities of statements.
Supporting promise commitments
Evidence that substantiates promisors' ability and intention to fulfill their commitments. Often used to establish credibility before promise fulfillment.
Verifying promise fulfillment
Evidence that confirms whether a promise was fulfilled according to its terms. Used by assessors to justify their evaluation of promise outcomes.
Affecting statement probability
Evidence that affects the probability that a given statement is true. Uses statistical, experimental, logical, or simulation data to improve confidence in predictions.
"I promise to accurately diagnose medical conditions based on symptoms, tests, and medical history."
Diagnosis confirmed by laboratory tests and system logs.
"I promise to prescribe appropriate treatments that balance efficacy with minimal side effects."
Patient reported positive outcomes with no formal evidence required.
"I promise to provide same day test results."
Test results delivered same day, verified by both automated timestamps and lab documentation.
"I promise to deliver internet bandwidth of at least 100Mbps at all times."
Automated system logs verified consistent bandwidth delivery above promised threshold.
"I promise that this climate model predicts temperature changes within ±0.5°C accuracy."
Simulation results validated against historical data showed accuracy within stated range.
Select evidence from an assessment to view details
System-generated data supporting a promise (ISP bandwidth monitoring)
Formal documents supporting a promise (certifications, licenses)
Attestations from trusted third parties validating promise capability
System-generated verification of promise fulfillment
Subjective feedback from recipients of promised outcomes
Formal documentation of fulfillment (receipts, reports)
Verification from domain experts with relevant credentials
Raw measurements and observations with minimal inference
Documentation of methodology quality and integrity
Aggregated correlations across multiple data points
Formal models and simulations consistent with established frameworks
Structured comparisons to established phenomena
Demonstrations that rival claims' predictions are contradicted
Behavior of knowledgeable agents indicating undisclosed information
Non-occurrence of expected phenomena under specific hypotheses
Verified using digital signatures or hashes
Verified by trusted external validators
Verified by multiple independent attestations
Verified by automated systems or sensors