P23 · CASE 0016 · POPULATED EVIDENCE RECORD
What Makes a Source Trustworthy?
OPENCOMPLETE_CLAIMS_SOURCES_ADVERSARIAL
Trustworthiness is a claim-specific, time-sensitive and revisable profile based on evidence access, method, provenance, independence, uncertainty, corrections and accountability.
17sources
15claims
10framework dimensions
15adversarial tests
Canonical source basis:
CP1_OBSERVATORY_P23_CASE0016_SOURCE_MAP_AND_EVIDENCE_MATRIX_V0_1_20260731.xlsx
Claims
Claims retain their original research-state labels. A status is not a probability of truth.
| ID | State | Claim | Support refs |
|---|---|---|---|
| C01 | PROVISIONAL SUPPORT | Trustworthiness is claim-specific rather than an all-purpose property of a source. | S02; S04; S08; S09; S11 |
| C02 | PROVISIONAL SUPPORT | Relevant expertise is useful but cannot replace inspectable evidence. | S04; S08; S09; S17 |
| C03 | PROVISIONAL SUPPORT | Transparent methods improve error detection and warranted reliance. | S04; S08; S15; S17 |
| C04 | STRONG SUPPORT | Digital provenance can establish origin and edit history but not truth. | S06; S07; S13; S14 |
| C05 | PROVISIONAL SUPPORT | Independent corroboration is stronger than repeated citation chains. | S04; S08; S09; S10 |
| C06 | PROVISIONAL SUPPORT | Visible corrections can be a positive trustworthiness signal. | S16; S17; S04 |
| C07 | PROVISIONAL SUPPORT | Institutional reputation reduces some risks but cannot validate every claim. | S04; S16; S17 |
| C08 | PROVISIONAL SUPPORT | Conflicts of interest reduce confidence when undisclosed; disclosure alone does not remove bias. | S08; S09; S10; S17 |
| C09 | PROVISIONAL SUPPORT | A large number of reports does not equal independent corroboration. | S04; S10; S13 |
| C10 | PROVISIONAL SUPPORT | Source trustworthiness changes with time, context and correction history. | S04; S09; S13; S17 |
| C11 | PROVISIONAL SUPPORT | AI fluency should not be treated as evidence of knowledge or verification. | S15; S04 |
| C12 | OPEN | No single universal source-trust score is currently justified. | S02–S17 |
| C13 | OPEN | Uncertainty disclosure is a positive signal only when calibrated to actual error. | S09; S15 |
| C14 | PROVISIONAL SUPPORT | Primary sources are not automatically superior to secondary synthesis. | S08; S09; S10 |
| C15 | PROVISIONAL SUPPORT | Trustworthiness can be increased by accountability and explainable decisions. | S15; S16; S17 |
Sources · 17 serialized
| ID | Source | Research status |
|---|---|---|
| S01 | Reuters Institute — Digital News Report 2026 | ACCEPTED CONTEXT |
| S02 | UNESCO — Verification of Information | ACCEPTED METHOD |
| S03 | UNESCO — MIL Module 4 | ACCEPTED METHOD |
| S04 | National Academies — Understanding and Addressing Misinformation About Science | ACCEPTED EVIDENCE |
| S05 | National Academies — Guide for Scientists to Identify and Address Misinformation | ACCEPTED METHOD |
| S06 | C2PA Technical Specification 2.4 | ACCEPTED TECHNICAL |
| S07 | C2PA FAQs | ACCEPTED CONTEXT |
| S08 | Cochrane Handbook Chapter 8 | ACCEPTED METHOD |
| S09 | Cochrane Handbook Chapter 14 / GRADE | ACCEPTED METHOD |
| S10 | Cochrane Handbook Chapter 13 | ACCEPTED METHOD |
| S11 | WHO — Global principles for credible health information sources | ACCEPTED METHOD |
| S12 | WHO — Social listening in infodemic management | ACCEPTED CONTEXT |
| S13 | Crossref — Research Nexus | ACCEPTED TECHNICAL |
| S14 | Crossref — Persistent identifiers in research infrastructure policy | ACCEPTED TECHNICAL |
| S15 | NIST — AI Risk Management Framework | ACCEPTED METHOD |
| S16 | Thomson Reuters Trust Principles | ACCEPTED CONTEXT |
| S17 | Society of Professional Journalists Code of Ethics | ACCEPTED METHOD |
Tests / adversarial controls · 15
| ID | Failure mode / test | State | Required control |
|---|---|---|---|
| A01 | Prestige laundering | ACTIVE | Return to claim-level evidence and method. |
| A02 | Citation laundering | ACTIVE | Map dependency and first origin. |
| A03 | Credential substitution | ACTIVE | Evaluate domain fit and evidence access. |
| A04 | Transparency theatre | ACTIVE | Identify decision-relevant method and data. |
| A05 | Provenance fallacy | ACTIVE | Separate origin verification from claim verification. |
| A06 | Balance fallacy | ACTIVE | Weight positions by evidence, not symmetry. |
| A07 | Recency fallacy | ACTIVE | Evaluate date, context and supersession. |
| A08 | Correction penalty paradox | ACTIVE | Score correction quality and preserved history. |
| A09 | AI fluency bias | ACTIVE | Require provenance, citations and claim verification. |
| A10 | Identity / affinity bias | ACTIVE | Blind claim review where feasible. |
| A11 | Primary-source absolutism | ACTIVE | Match source type to question and assess bias. |
| A12 | Consensus shortcut | ACTIVE | Inspect how consensus was produced and bounded. |
| A13 | Conflict absolutism | ACTIVE | Record conflict and test observable effect. |
| A14 | Metric gaming | ACTIVE | Keep hard gates and qualitative review. |
| A15 | Context stripping | ACTIVE | Inspect surrounding context and transformations. |
Serialization boundary
This interface does not complete missing research by inference. The current record is COMPLETE_CLAIMS_SOURCES_ADVERSARIAL. Missing source-level or bilingual object-level fields remain absent until they are read from a canonical Observatory artifact.