Changes for page FAQ
Last modified by Robert Schaub on 2025/12/24 20:33
Summary
-
Page properties (1 modified, 0 added, 0 removed)
Details
- Page properties
-
- Content
-
... ... @@ -13,6 +13,7 @@ 13 13 **What**: System dynamically researches claims using AKEL (AI Knowledge Extraction Layer) 14 14 15 15 **Process**: 16 + 16 16 * Extracts claims from submitted text 17 17 * Generates structured sub-queries 18 18 * Performs **mandatory contradiction search** (actively seeks counter-evidence, not just confirmations) ... ... @@ -40,6 +40,7 @@ 40 40 **What**: Sampling audits where experts review AI-generated content 41 41 42 42 **Rates**: 44 + 43 43 * High-risk (Tier A): 30-50% sampling 44 44 * Medium-risk (Tier B): 10-20% sampling 45 45 * Low-risk (Tier C): 5-10% sampling ... ... @@ -51,6 +51,7 @@ 51 51 === Why All Three Matter === 52 52 53 53 **Complementary Strengths**: 56 + 54 54 * **AI research**: Scale and speed for emerging claims 55 55 * **Expert authoring**: Authority and precision for critical domains 56 56 * **Audit feedback**: Continuous quality improvement ... ... @@ -58,6 +58,7 @@ 58 58 **Expert Time Optimization**: 59 59 60 60 Experts can choose where to focus their time: 64 + 61 61 * Author high-priority content directly 62 62 * Validate and edit AI-generated outputs 63 63 * Audit samples to improve system-wide AI performance ... ... @@ -77,6 +77,7 @@ 77 77 FactHarbor includes multiple safeguards against echo chambers and filter bubbles: 78 78 79 79 **Mandatory Contradiction Search**: 84 + 80 80 * AI must actively search for counter-evidence, not just confirmations 81 81 * System checks for echo chamber patterns in source clusters 82 82 * Flags tribal or ideological source clustering ... ... @@ -83,21 +83,25 @@ 83 83 * Requires diverse perspectives across political/ideological spectrum 84 84 85 85 **Multiple Scenarios**: 91 + 86 86 * Claims are evaluated under different interpretations 87 87 * Reveals how assumptions change conclusions 88 88 * Makes disagreements understandable, not divisive 89 89 90 90 **Transparent Reasoning**: 97 + 91 91 * All assumptions, definitions, and boundaries are explicit 92 92 * Evidence chains are traceable 93 93 * Uncertainty is quantified, not hidden 94 94 95 95 **Audit System**: 103 + 96 96 * Human auditors check for bubble patterns 97 97 * Feedback loop improves AI search diversity 98 98 * Community can flag missing perspectives 99 99 100 100 **Federation**: 109 + 101 101 * Multiple independent nodes with different perspectives 102 102 * No single entity controls "the truth" 103 103 * Cross-node contradiction detection ... ... @@ -109,20 +109,23 @@ 109 109 This is exactly what FactHarbor is designed for: 110 110 111 111 **Scenarios capture contexts**: 121 + 112 112 * Each scenario defines specific boundaries, definitions, and assumptions 113 113 * The same claim can have different verdicts in different scenarios 114 114 * Example: "Coffee is healthy" depends on: 115 - 116 - 117 - 118 - 125 +** Definition of "healthy" (reduces disease risk? improves mood? affects specific conditions?) 126 +** Population (adults? pregnant women? people with heart conditions?) 127 +** Consumption level (1 cup/day? 5 cups/day?) 128 +** Time horizon (short-term? long-term?) 119 119 120 120 **Truth Landscape**: 131 + 121 121 * Shows all scenarios and their verdicts side-by-side 122 122 * Users see *why* interpretations differ 123 123 * No forced consensus when legitimate disagreement exists 124 124 125 125 **Explicit Assumptions**: 137 + 126 126 * Every scenario states its assumptions clearly 127 127 * Users can compare how changing assumptions changes conclusions 128 128 * Makes context-dependence visible, not hidden ... ... @@ -132,6 +132,7 @@ 132 132 == What makes FactHarbor different from traditional fact-checking sites? == 133 133 134 134 **Traditional Fact-Checking**: 147 + 135 135 * Binary verdicts: True / Mostly True / False 136 136 * Single interpretation chosen by fact-checker 137 137 * Often hides legitimate contextual differences ... ... @@ -138,6 +138,7 @@ 138 138 * Limited ability to show *why* people disagree 139 139 140 140 **FactHarbor**: 154 + 141 141 * **Multi-scenario**: Shows multiple valid interpretations 142 142 * **Likelihood-based**: Ranges with uncertainty, not binary labels 143 143 * **Transparent assumptions**: Makes boundaries and definitions explicit ... ... @@ -150,6 +150,7 @@ 150 150 == How do you prevent manipulation or coordinated misinformation campaigns? == 151 151 152 152 **Quality Gates**: 167 + 153 153 * Automated checks before AI-generated content publishes 154 154 * Source quality verification 155 155 * Mandatory contradiction search ... ... @@ -156,11 +156,13 @@ 156 156 * Bubble detection for coordinated campaigns 157 157 158 158 **Audit System**: 174 + 159 159 * Stratified sampling catches manipulation patterns 160 160 * Expert auditors validate AI research quality 161 161 * Failed audits trigger immediate review 162 162 163 163 **Transparency**: 180 + 164 164 * All reasoning chains are visible 165 165 * Evidence sources are traceable 166 166 * AKEL involvement clearly labeled ... ... @@ -167,11 +167,13 @@ 167 167 * Version history preserved 168 168 169 169 **Moderation**: 187 + 170 170 * Moderators handle abuse, spam, coordinated manipulation 171 171 * Content can be flagged by community 172 172 * Audit trail maintained even if content hidden 173 173 174 174 **Federation**: 193 + 175 175 * Multiple nodes with independent governance 176 176 * No single point of control 177 177 * Cross-node contradiction detection ... ... @@ -184,6 +184,7 @@ 184 184 FactHarbor is designed for evolving knowledge: 185 185 186 186 **Automatic Re-evaluation**: 206 + 187 187 1. New evidence arrives 188 188 2. System detects affected scenarios and verdicts 189 189 3. AKEL proposes updated verdicts ... ... @@ -192,16 +192,19 @@ 192 192 6. Old versions remain accessible 193 193 194 194 **Version History**: 215 + 195 195 * Every verdict has complete history 196 196 * Users can see "as of date X, what did we know?" 197 197 * Timeline shows how understanding evolved 198 198 199 199 **Transparent Updates**: 221 + 200 200 * Reason for re-evaluation documented 201 201 * New evidence clearly linked 202 202 * Changes explained, not hidden 203 203 204 204 **User Notifications**: 227 + 205 205 * Users following claims are notified of updates 206 206 * Can compare old vs new verdicts 207 207 * Can see which evidence changed conclusions ... ... @@ -213,6 +213,7 @@ 213 213 **Anyone** - even without login: 214 214 215 215 **Readers** (no login required): 239 + 216 216 * Browse and search all published content 217 217 * Submit text for analysis 218 218 * New claims added automatically unless duplicates exist ... ... @@ -219,6 +219,7 @@ 219 219 * System deduplicates and normalizes 220 220 221 221 **Contributors** (logged in): 246 + 222 222 * Everything Readers can do 223 223 * Submit evidence sources 224 224 * Suggest scenarios ... ... @@ -225,6 +225,7 @@ 225 225 * Participate in discussions 226 226 227 227 **Workflow**: 253 + 228 228 1. User submits text (as Reader or Contributor) 229 229 2. AKEL extracts claims 230 230 3. Checks for existing duplicates ... ... @@ -241,6 +241,7 @@ 241 241 Risk tiers determine review requirements and publication workflow: 242 242 243 243 **Tier A (High Risk)**: 270 + 244 244 * **Domains**: Medical, legal, elections, safety, security, major financial 245 245 * **Publication**: AI can publish with warnings, expert review required for "Human-Reviewed" status 246 246 * **Audit rate**: Recommendation 30-50% ... ... @@ -247,6 +247,7 @@ 247 247 * **Why**: Potential for significant harm if wrong 248 248 249 249 **Tier B (Medium Risk)**: 277 + 250 250 * **Domains**: Complex policy, science causality, contested issues 251 251 * **Publication**: AI can publish immediately with clear labeling 252 252 * **Audit rate**: Recommendation 10-20% ... ... @@ -253,6 +253,7 @@ 253 253 * **Why**: Nuanced but lower immediate harm risk 254 254 255 255 **Tier C (Low Risk)**: 284 + 256 256 * **Domains**: Definitions, established facts, historical data 257 257 * **Publication**: AI publication default 258 258 * **Audit rate**: Recommendation 5-10% ... ... @@ -259,6 +259,7 @@ 259 259 * **Why**: Well-established, low controversy 260 260 261 261 **Assignment**: 291 + 262 262 * AKEL suggests tier based on domain, keywords, impact 263 263 * Moderators and Experts can override 264 264 * Risk tiers reviewed based on audit outcomes ... ... @@ -268,6 +268,7 @@ 268 268 == How does federation work and why is it important? == 269 269 270 270 **Federation Model**: 301 + 271 271 * Multiple independent FactHarbor nodes 272 272 * Each node has own database, AKEL, governance 273 273 * Nodes exchange claims, scenarios, evidence, verdicts ... ... @@ -274,6 +274,7 @@ 274 274 * No central authority 275 275 276 276 **Why Federation Matters**: 308 + 277 277 * **Resilience**: No single point of failure or censorship 278 278 * **Autonomy**: Communities govern themselves 279 279 * **Scalability**: Add nodes to handle more users ... ... @@ -281,6 +281,7 @@ 281 281 * **Trust diversity**: Multiple perspectives, not single truth source 282 282 283 283 **How Nodes Exchange Data**: 316 + 284 284 1. Local node creates versions 285 285 2. Builds signed bundle 286 286 3. Pushes to trusted neighbor nodes ... ... @@ -289,6 +289,7 @@ 289 289 6. Local re-evaluation if needed 290 290 291 291 **Trust Model**: 325 + 292 292 * Trusted nodes → auto-import 293 293 * Neutral nodes → import with review 294 294 * Untrusted nodes → manual only ... ... @@ -300,16 +300,19 @@ 300 300 **Yes - and that's a feature, not a bug**: 301 301 302 302 **Multiple Scenarios**: 337 + 303 303 * Experts can create different scenarios with different assumptions 304 304 * Each scenario gets its own verdict 305 305 * Users see *why* experts disagree (different definitions, boundaries, evidence weighting) 306 306 307 307 **Parallel Verdicts**: 343 + 308 308 * Same scenario, different expert interpretations 309 309 * Both verdicts visible with expert attribution 310 310 * No forced consensus 311 311 312 312 **Transparency**: 349 + 313 313 * Expert reasoning documented 314 314 * Assumptions stated explicitly 315 315 * Evidence chains traceable ... ... @@ -316,6 +316,7 @@ 316 316 * Users can evaluate competing expert opinions 317 317 318 318 **Federation**: 356 + 319 319 * Different nodes can have different expert conclusions 320 320 * Cross-node branching allowed 321 321 * Users can see how conclusions vary across nodes ... ... @@ -327,6 +327,7 @@ 327 327 **Multiple Safeguards**: 328 328 329 329 **Quality Gate 4: Structural Integrity**: 368 + 330 330 * Fact-checking against sources 331 331 * No hallucinations allowed 332 332 * Logic chain must be valid and traceable ... ... @@ -333,6 +333,7 @@ 333 333 * References must be accessible and verifiable 334 334 335 335 **Evidence Requirements**: 375 + 336 336 * Primary sources required 337 337 * Citations must be complete 338 338 * Sources must be accessible ... ... @@ -339,11 +339,13 @@ 339 339 * Reliability scored 340 340 341 341 **Audit System**: 382 + 342 342 * Human auditors check AI-generated content 343 343 * Hallucinations caught and fed back into training 344 344 * Patterns of errors trigger system improvements 345 345 346 346 **Transparency**: 388 + 347 347 * All reasoning chains visible 348 348 * Sources linked 349 349 * Users can verify claims against sources ... ... @@ -350,6 +350,7 @@ 350 350 * AKEL outputs clearly labeled 351 351 352 352 **Human Oversight**: 395 + 353 353 * Tier A requires expert review for "Human-Reviewed" status 354 354 * Audit sampling catches errors 355 355 * Community can flag issues ... ... @@ -361,6 +361,7 @@ 361 361 [ToDo: Business model and sustainability to be defined] 362 362 363 363 Potential models under consideration: 407 + 364 364 * Non-profit foundation with grants and donations 365 365 * Institutional subscriptions (universities, research organizations, media) 366 366 * API access for third-party integrations ... ... @@ -377,5 +377,4 @@ 377 377 * [[AKEL (AI Knowledge Extraction Layer)>>FactHarbor.Specification.AI Knowledge Extraction Layer (AKEL).WebHome]] 378 378 * [[Automation>>FactHarbor.Specification.Automation.WebHome]] 379 379 * [[Federation & Decentralization>>FactHarbor.Specification.Federation & Decentralization.WebHome]] 380 -* [[Mission & Purpose>>FactHarbor.Organisation.Core Problems FactHarbor Solves.WebHome]] 381 - 424 +* [[Mission & Purpose>>FactHarbor.Organisation V0\.9\.18.Core Problems FactHarbor Solves.WebHome]]