Skip to Content
Last modified by Robert Schaub on 2025/12/24 20:33
From version 6.4
edited by Robert Schaub
on 2025/12/16 21:39
Change comment: Renamed back-links.
To version 5.1
edited by Robert Schaub
on 2025/12/14 18:59
Change comment: Imported from XAR

Summary

Details

Page properties
Parent
... ... @@ -1,1 +1,1 @@
1 -FactHarbor.Specification V0\.9\.18.WebHome
1 +FactHarbor.Specification.WebHome
Content
... ... @@ -50,7 +50,6 @@
50 50  == Inputs and Outputs ==
51 51  
52 52  === Inputs ===
53 -
54 54  * User-submitted claims or evidence
55 55  * Uploaded documents
56 56  * URLs or citations
... ... @@ -58,7 +58,6 @@
58 58  * Embeddings (from local or federated peers)
59 59  
60 60  === Outputs (publication mode varies by risk tier) ===
61 -
62 62  * ClaimVersion (draft or AI-generated)
63 63  * ScenarioVersion (draft or AI-generated)
64 64  * EvidenceVersion (summary + metadata, draft or AI-generated)
... ... @@ -77,7 +77,6 @@
77 77  === Mode 1: Draft-Only (Never Public) ===
78 78  
79 79  **Used for:**
80 -
81 81  * Failed quality gate checks
82 82  * Sensitive topics flagged for expert review
83 83  * Unclear scope or missing critical sources
... ... @@ -88,7 +88,6 @@
88 88  === Mode 2: Published as AI-Generated (No Prior Human Review) ===
89 89  
90 90  **Requirements:**
91 -
92 92  * All automated quality gates passed (see below)
93 93  * Risk tier permits AI-draft publication (Tier B or C)
94 94  * Contradiction search completed successfully
... ... @@ -102,7 +102,6 @@
102 102  ```
103 103  
104 104  **User actions:**
105 -
106 106  * Browse and read content
107 107  * Request human review (escalates to review queue)
108 108  * Flag for expert attention
... ... @@ -110,7 +110,6 @@
110 110  === Mode 3: Published as Human-Reviewed ===
111 111  
112 112  **Requirements:**
113 -
114 114  * Human reviewer or domain expert validated
115 115  * All quality gates passed
116 116  * Visible "Human-Reviewed" mark with reviewer role and timestamp
... ... @@ -133,13 +133,12 @@
133 133  **Domains:** Medical, legal, elections, safety/security, major reputational harm
134 134  
135 135  **Publication policy:**
136 -
137 137  * Human review REQUIRED before "Human-Reviewed" status
138 138  * AI-generated content MAY be published but:
139 -** Clearly flagged as AI-draft with prominent disclaimer
140 -** May have limited visibility
141 -** Auto-escalated to expert review queue
142 -** User warnings displayed
132 + ** Clearly flagged as AI-draft with prominent disclaimer
133 + ** May have limited visibility
134 + ** Auto-escalated to expert review queue
135 + ** User warnings displayed
143 143  
144 144  **Audit rate:** Recommendation: 30-50% of published AI-drafts sampled in first 6 months
145 145  
... ... @@ -148,7 +148,6 @@
148 148  **Domains:** Contested public policy, complex science, causality claims, significant financial impact
149 149  
150 150  **Publication policy:**
151 -
152 152  * AI-draft CAN publish immediately with clear labeling
153 153  * Sampling audits conducted (see Audit System below)
154 154  * High-engagement items auto-escalated to expert review
... ... @@ -161,7 +161,6 @@
161 161  **Domains:** Definitions, simple factual lookups with strong primary sources, historical facts, established scientific consensus
162 162  
163 163  **Publication policy:**
164 -
165 165  * AI-draft default publication mode
166 166  * Sampling audits sufficient
167 167  * Community flagging available
... ... @@ -176,7 +176,6 @@
176 176  All AI-generated content must pass these automated checks before Mode 2 publication:
177 177  
178 178  === Gate 1: Source Quality ===
179 -
180 180  * Primary sources identified and accessible
181 181  * Source reliability scored against whitelist
182 182  * Citation completeness verified
... ... @@ -193,7 +193,6 @@
193 193  * **Bubble detection** – Conspiracy theories, echo chambers, ideologically isolated sources
194 194  
195 195  **Search coverage requirements:**
196 -
197 197  * Academic literature (BOTH supporting AND opposing views)
198 198  * Reputable media across diverse political/ideological perspectives
199 199  * Official contradictions (retractions, corrections, updates, amendments)
... ... @@ -201,7 +201,6 @@
201 201  * Cross-cultural and international perspectives
202 202  
203 203  **Search must actively avoid algorithmic bubbles:**
204 -
205 205  * Deliberately seek opposing viewpoints
206 206  * Check for echo chamber patterns in source clusters
207 207  * Identify tribal or ideological source clustering
... ... @@ -209,7 +209,6 @@
209 209  * Verify diversity of perspectives represented
210 210  
211 211  **Outcomes:**
212 -
213 213  * **Strong counter-evidence found** → Auto-escalate to Tier B or draft-only mode
214 214  * **Significant uncertainty detected** → Require uncertainty disclosure in verdict
215 215  * **Bubble indicators present** → Flag for expert review and human validation
... ... @@ -216,7 +216,6 @@
216 216  * **Limited perspective diversity** → Expand search or flag for human review
217 217  
218 218  === Gate 3: Uncertainty Quantification ===
219 -
220 220  * Confidence scores calculated for all claims and verdicts
221 221  * Limitations explicitly stated
222 222  * Data gaps identified and disclosed
... ... @@ -224,7 +224,6 @@
224 224  * Alternative scenarios considered
225 225  
226 226  === Gate 4: Structural Integrity ===
227 -
228 228  * No hallucinations detected (fact-checking against sources)
229 229  * Logic chain valid and traceable
230 230  * References accessible and verifiable
... ... @@ -232,7 +232,6 @@
232 232  * Premises clearly stated
233 233  
234 234  **If any gate fails:**
235 -
236 236  * Content remains in draft-only mode
237 237  * Failure reason logged
238 238  * Human review required before publication
... ... @@ -247,7 +247,6 @@
247 247  === Sampling Strategy ===
248 248  
249 249  Audits prioritize:
250 -
251 251  * **Risk tier** (higher tiers get more frequent audits)
252 252  * **AI confidence score** (low confidence → higher sampling rate)
253 253  * **Traffic and engagement** (high-visibility content audited more)
... ... @@ -259,12 +259,11 @@
259 259  1. System selects content for audit based on sampling strategy
260 260  2. Human auditor reviews AI-generated content against quality standards
261 261  3. Auditor validates or corrects:
262 -
263 -* Claim extraction accuracy
264 -* Scenario appropriateness
265 -* Evidence relevance and interpretation
266 -* Verdict reasoning
267 -* Contradiction search completeness
245 + * Claim extraction accuracy
246 + * Scenario appropriateness
247 + * Evidence relevance and interpretation
248 + * Verdict reasoning
249 + * Contradiction search completeness
268 268  4. Audit outcome recorded (pass/fail + detailed feedback)
269 269  5. Failed audits trigger immediate content review
270 270  6. Audit results feed back into system improvement
... ... @@ -272,7 +272,6 @@
272 272  === Feedback Loop (Continuous Improvement) ===
273 273  
274 274  Audit outcomes systematically improve:
275 -
276 276  * **Query templates** – Refined based on missed evidence patterns
277 277  * **Retrieval source weights** – Adjusted for accuracy and reliability
278 278  * **Contradiction detection heuristics** – Enhanced to catch missed counter-evidence
... ... @@ -291,7 +291,7 @@
291 291  
292 292  == Architecture Overview ==
293 293  
294 -{{include reference="FactHarbor.Archive.FactHarbor V0\.9\.23 Lost Data.Specification.Diagrams.AKEL Architecture.WebHome"/}}
275 +{{include reference="FactHarbor.Specification.Diagrams.AKEL Architecture.WebHome"/}}
295 295  
296 296  ----
297 297  
... ... @@ -346,7 +346,8 @@
346 346  
347 347  == Related Pages ==
348 348  
349 -* [[Automation>>FactHarbor.Specification V0\.9\.18.Automation.WebHome]]
330 +* [[Automation>>FactHarbor.Specification.Automation.WebHome]]
350 350  * [[Requirements (Roles)>>FactHarbor.Specification.Requirements.WebHome]]
351 351  * [[Workflows>>FactHarbor.Specification.Workflows.WebHome]]
352 352  * [[Governance>>FactHarbor.Organisation.Governance]]
334 +