Wiki source code of FAQ

Version 2.1 by Robert Schaub on 2025/12/14 23:02

Show last authors
1 = Frequently Asked Questions (FAQ) =
2
3 Common questions about FactHarbor's design, functionality, and approach.
4
5 ----
6
7 == How do facts get input into the system? ==
8
9 FactHarbor uses a hybrid model:
10
11 **~1. **AI-Generated (scalable)**: System dynamically researches claims—extracting, generating structured sub-queries, performing mandatory contradiction search (actively seeking counter-evidence, not just confirmations), running quality gates. Published with clear "AI-Generated" labels.**
12
13 **2. Expert-Authored (authoritative)**: Domain experts directly author, edit, and validate content—especially for high-risk domains (medical, legal). These get "Human-Reviewed" status and higher trust.
14
15 **3. Audit-Improved (continuous quality)**: Sampling audits (30-50% high-risk, 5-10% low-risk) where expert reviews systematically improve AI research quality.
16
17 **Why both matter**:
18
19 * AI research handles scale—emerging claims, immediate responses with transparent reasoning
20 * Expert authoring provides authoritative grounding for critical domains
21 * Audit feedback ensures AI quality improves based on expert validation patterns
22
23 Experts can author high-priority content directly, validate/edit AI outputs, or audit samples to improve system-wide performance—focusing their time where expertise matters most.
24
25 POC v1 demonstrates the AI research pipeline (fully automated with transparent reasoning); full system supports all three pathways.
26
27 ----
28
29 == What prevents FactHarbor from becoming another echo chamber? ==
30
31 FactHarbor includes multiple safeguards against echo chambers and filter bubbles:
32
33 **Mandatory Contradiction Search**:
34
35 * AI must actively search for counter-evidence, not just confirmations
36 * System checks for echo chamber patterns in source clusters
37 * Flags tribal or ideological source clustering
38 * Requires diverse perspectives across political/ideological spectrum
39
40 **Multiple Scenarios**:
41
42 * Claims are evaluated under different interpretations
43 * Reveals how assumptions change conclusions
44 * Makes disagreements understandable, not divisive
45
46 **Transparent Reasoning**:
47
48 * All assumptions, definitions, and boundaries are explicit
49 * Evidence chains are traceable
50 * Uncertainty is quantified, not hidden
51
52 **Audit System**:
53
54 * Human auditors check for bubble patterns
55 * Feedback loop improves AI search diversity
56 * Community can flag missing perspectives
57
58 **Federation**:
59
60 * Multiple independent nodes with different perspectives
61 * No single entity controls "the truth"
62 * Cross-node contradiction detection
63
64 ----
65
66 == How does FactHarbor handle claims that are "true in one context but false in another"? ==
67
68 This is exactly what FactHarbor is designed for:
69
70 **Scenarios capture contexts**:
71
72 * Each scenario defines specific boundaries, definitions, and assumptions
73 * The same claim can have different verdicts in different scenarios
74 * Example: "Coffee is healthy" depends on:
75 ** Definition of "healthy" (reduces disease risk? improves mood? affects specific conditions?)
76 ** Population (adults? pregnant women? people with heart conditions?)
77 ** Consumption level (1 cup/day? 5 cups/day?)
78 ** Time horizon (short-term? long-term?)
79
80 **Truth Landscape**:
81
82 * Shows all scenarios and their verdicts side-by-side
83 * Users see *why* interpretations differ
84 * No forced consensus when legitimate disagreement exists
85
86 **Explicit Assumptions**:
87
88 * Every scenario states its assumptions clearly
89 * Users can compare how changing assumptions changes conclusions
90 * Makes context-dependence visible, not hidden
91
92 ----
93
94 == What makes FactHarbor different from traditional fact-checking sites? ==
95
96 **Traditional Fact-Checking**:
97
98 * Binary verdicts: True / Mostly True / False
99 * Single interpretation chosen by fact-checker
100 * Often hides legitimate contextual differences
101 * Limited ability to show *why* people disagree
102
103 **FactHarbor**:
104
105 * **Multi-scenario**: Shows multiple valid interpretations
106 * **Likelihood-based**: Ranges with uncertainty, not binary labels
107 * **Transparent assumptions**: Makes boundaries and definitions explicit
108 * **Version history**: Shows how understanding evolves
109 * **Contradiction search**: Actively seeks opposing evidence
110 * **Federated**: No single authority controls truth
111
112 ----
113
114 == How do you prevent manipulation or coordinated misinformation campaigns? ==
115
116 **Quality Gates**:
117
118 * Automated checks before AI-generated content publishes
119 * Source quality verification
120 * Mandatory contradiction search
121 * Bubble detection for coordinated campaigns
122
123 **Audit System**:
124
125 * Stratified sampling catches manipulation patterns
126 * Expert auditors validate AI research quality
127 * Failed audits trigger immediate review
128
129 **Transparency**:
130
131 * All reasoning chains are visible
132 * Evidence sources are traceable
133 * AKEL involvement clearly labeled
134 * Version history preserved
135
136 **Moderation**:
137
138 * Moderators handle abuse, spam, coordinated manipulation
139 * Content can be flagged by community
140 * Audit trail maintained even if content hidden
141
142 **Federation**:
143
144 * Multiple nodes with independent governance
145 * No single point of control
146 * Cross-node contradiction detection
147 * Trust model prevents malicious node influence
148
149 ----
150
151 == What happens when new evidence contradicts an existing verdict? ==
152
153 FactHarbor is designed for evolving knowledge:
154
155 **Automatic Re-evaluation**:
156
157 1. New evidence arrives
158 2. System detects affected scenarios and verdicts
159 3. AKEL proposes updated verdicts
160 4. Reviewers/experts validate
161 5. New verdict version published
162 6. Old versions remain accessible
163
164 **Version History**:
165
166 * Every verdict has complete history
167 * Users can see "as of date X, what did we know?"
168 * Timeline shows how understanding evolved
169
170 **Transparent Updates**:
171
172 * Reason for re-evaluation documented
173 * New evidence clearly linked
174 * Changes explained, not hidden
175
176 **User Notifications**:
177
178 * Users following claims are notified of updates
179 * Can compare old vs new verdicts
180 * Can see which evidence changed conclusions
181
182 ----
183
184 == Who can submit claims to FactHarbor? ==
185
186 **Anyone** - even without login:
187
188 **Readers** (no login required):
189
190 * Browse and search all published content
191 * Submit text for analysis
192 * New claims added automatically unless duplicates exist
193 * System deduplicates and normalizes
194
195 **Contributors** (logged in):
196
197 * Everything Readers can do
198 * Submit evidence sources
199 * Suggest scenarios
200 * Participate in discussions
201
202 **Workflow**:
203
204 1. User submits text (as Reader or Contributor)
205 2. AKEL extracts claims
206 3. Checks for existing duplicates
207 4. Normalizes claim text
208 5. Assigns risk tier
209 6. Generates scenarios (draft)
210 7. Runs quality gates
211 8. Publishes as AI-Generated (Mode 2) if passes
212
213 ----
214
215 == What are "risk tiers" and why do they matter? ==
216
217 Risk tiers determine review requirements and publication workflow:
218
219 **Tier A (High Risk)**:
220
221 * **Domains**: Medical, legal, elections, safety, security, major financial
222 * **Publication**: AI can publish with warnings, expert review required for "Human-Reviewed" status
223 * **Audit rate**: Recommendation 30-50%
224 * **Why**: Potential for significant harm if wrong
225
226 **Tier B (Medium Risk)**:
227
228 * **Domains**: Complex policy, science causality, contested issues
229 * **Publication**: AI can publish immediately with clear labeling
230 * **Audit rate**: Recommendation 10-20%
231 * **Why**: Nuanced but lower immediate harm risk
232
233 **Tier C (Low Risk)**:
234
235 * **Domains**: Definitions, established facts, historical data
236 * **Publication**: AI publication default
237 * **Audit rate**: Recommendation 5-10%
238 * **Why**: Well-established, low controversy
239
240 **Assignment**:
241
242 * AKEL suggests tier based on domain, keywords, impact
243 * Moderators and Experts can override
244 * Risk tiers reviewed based on audit outcomes
245
246 ----
247
248 == How does federation work and why is it important? ==
249
250 **Federation Model**:
251
252 * Multiple independent FactHarbor nodes
253 * Each node has own database, AKEL, governance
254 * Nodes exchange claims, scenarios, evidence, verdicts
255 * No central authority
256
257 **Why Federation Matters**:
258
259 * **Resilience**: No single point of failure or censorship
260 * **Autonomy**: Communities govern themselves
261 * **Scalability**: Add nodes to handle more users
262 * **Specialization**: Domain-focused nodes (health, energy, etc.)
263 * **Trust diversity**: Multiple perspectives, not single truth source
264
265 **How Nodes Exchange Data**:
266
267 1. Local node creates versions
268 2. Builds signed bundle
269 3. Pushes to trusted neighbor nodes
270 4. Remote nodes validate signatures and lineage
271 5. Accept or branch versions
272 6. Local re-evaluation if needed
273
274 **Trust Model**:
275
276 * Trusted nodes → auto-import
277 * Neutral nodes → import with review
278 * Untrusted nodes → manual only
279
280 ----
281
282 == Can experts disagree in FactHarbor? ==
283
284 **Yes - and that's a feature, not a bug**:
285
286 **Multiple Scenarios**:
287
288 * Experts can create different scenarios with different assumptions
289 * Each scenario gets its own verdict
290 * Users see *why* experts disagree (different definitions, boundaries, evidence weighting)
291
292 **Parallel Verdicts**:
293
294 * Same scenario, different expert interpretations
295 * Both verdicts visible with expert attribution
296 * No forced consensus
297
298 **Transparency**:
299
300 * Expert reasoning documented
301 * Assumptions stated explicitly
302 * Evidence chains traceable
303 * Users can evaluate competing expert opinions
304
305 **Federation**:
306
307 * Different nodes can have different expert conclusions
308 * Cross-node branching allowed
309 * Users can see how conclusions vary across nodes
310
311 ----
312
313 == What prevents AI from hallucinating or making up facts? ==
314
315 **Multiple Safeguards**:
316
317 **Quality Gate 4: Structural Integrity**:
318
319 * Fact-checking against sources
320 * No hallucinations allowed
321 * Logic chain must be valid and traceable
322 * References must be accessible and verifiable
323
324 **Evidence Requirements**:
325
326 * Primary sources required
327 * Citations must be complete
328 * Sources must be accessible
329 * Reliability scored
330
331 **Audit System**:
332
333 * Human auditors check AI-generated content
334 * Hallucinations caught and fed back into training
335 * Patterns of errors trigger system improvements
336
337 **Transparency**:
338
339 * All reasoning chains visible
340 * Sources linked
341 * Users can verify claims against sources
342 * AKEL outputs clearly labeled
343
344 **Human Oversight**:
345
346 * Tier A requires expert review for "Human-Reviewed" status
347 * Audit sampling catches errors
348 * Community can flag issues
349
350 ----
351
352 == How does FactHarbor make money / is it sustainable? ==
353
354 [ToDo: Business model and sustainability to be defined]
355
356 Potential models under consideration:
357
358 * Non-profit foundation with grants and donations
359 * Institutional subscriptions (universities, research organizations, media)
360 * API access for third-party integrations
361 * Premium features for power users
362 * Federated node hosting services
363
364 Core principle: **Public benefit** mission takes priority over profit.
365
366 ----
367
368 == Related Pages ==
369
370 * [[Requirements (Roles)>>FactHarbor.Specification.Requirements.WebHome]]
371 * [[AKEL (AI Knowledge Extraction Layer)>>FactHarbor.Specification.AI Knowledge Extraction Layer (AKEL).WebHome]]
372 * [[Automation>>FactHarbor.Specification.Automation.WebHome]]
373 * [[Federation & Decentralization>>FactHarbor.Specification.Federation & Decentralization.WebHome]]
374 * [[Mission & Purpose>>FactHarbor.Organisation.Mission & Purpose.WebHome]]