Skip to Content

Wiki source code of FAQ

Version 1.1 by Robert Schaub on 2025/12/14 22:27
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 * AI research handles scale—emerging claims, immediate responses with transparent reasoning
19 * Expert authoring provides authoritative grounding for critical domains
20 * Audit feedback ensures AI quality improves based on expert validation patterns
21
22 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.
23
24 POC v1 demonstrates the AI research pipeline (fully automated with transparent reasoning); full system supports all three pathways.
25
26 ----
27
28 == What prevents FactHarbor from becoming another echo chamber? ==
29
30 FactHarbor includes multiple safeguards against echo chambers and filter bubbles:
31
32 **Mandatory Contradiction Search**:
33 * AI must actively search for counter-evidence, not just confirmations
34 * System checks for echo chamber patterns in source clusters
35 * Flags tribal or ideological source clustering
36 * Requires diverse perspectives across political/ideological spectrum
37
38 **Multiple Scenarios**:
39 * Claims are evaluated under different interpretations
40 * Reveals how assumptions change conclusions
41 * Makes disagreements understandable, not divisive
42
43 **Transparent Reasoning**:
44 * All assumptions, definitions, and boundaries are explicit
45 * Evidence chains are traceable
46 * Uncertainty is quantified, not hidden
47
48 **Audit System**:
49 * Human auditors check for bubble patterns
50 * Feedback loop improves AI search diversity
51 * Community can flag missing perspectives
52
53 **Federation**:
54 * Multiple independent nodes with different perspectives
55 * No single entity controls "the truth"
56 * Cross-node contradiction detection
57
58 ----
59
60 == How does FactHarbor handle claims that are "true in one context but false in another"? ==
61
62 This is exactly what FactHarbor is designed for:
63
64 **Scenarios capture contexts**:
65 * Each scenario defines specific boundaries, definitions, and assumptions
66 * The same claim can have different verdicts in different scenarios
67 * Example: "Coffee is healthy" depends on:
68 ** Definition of "healthy" (reduces disease risk? improves mood? affects specific conditions?)
69 ** Population (adults? pregnant women? people with heart conditions?)
70 ** Consumption level (1 cup/day? 5 cups/day?)
71 ** Time horizon (short-term? long-term?)
72
73 **Truth Landscape**:
74 * Shows all scenarios and their verdicts side-by-side
75 * Users see *why* interpretations differ
76 * No forced consensus when legitimate disagreement exists
77
78 **Explicit Assumptions**:
79 * Every scenario states its assumptions clearly
80 * Users can compare how changing assumptions changes conclusions
81 * Makes context-dependence visible, not hidden
82
83 ----
84
85 == What makes FactHarbor different from traditional fact-checking sites? ==
86
87 **Traditional Fact-Checking**:
88 * Binary verdicts: True / Mostly True / False
89 * Single interpretation chosen by fact-checker
90 * Often hides legitimate contextual differences
91 * Limited ability to show *why* people disagree
92
93 **FactHarbor**:
94 * **Multi-scenario**: Shows multiple valid interpretations
95 * **Likelihood-based**: Ranges with uncertainty, not binary labels
96 * **Transparent assumptions**: Makes boundaries and definitions explicit
97 * **Version history**: Shows how understanding evolves
98 * **Contradiction search**: Actively seeks opposing evidence
99 * **Federated**: No single authority controls truth
100
101 ----
102
103 == How do you prevent manipulation or coordinated misinformation campaigns? ==
104
105 **Quality Gates**:
106 * Automated checks before AI-generated content publishes
107 * Source quality verification
108 * Mandatory contradiction search
109 * Bubble detection for coordinated campaigns
110
111 **Audit System**:
112 * Stratified sampling catches manipulation patterns
113 * Expert auditors validate AI research quality
114 * Failed audits trigger immediate review
115
116 **Transparency**:
117 * All reasoning chains are visible
118 * Evidence sources are traceable
119 * AKEL involvement clearly labeled
120 * Version history preserved
121
122 **Moderation**:
123 * Moderators handle abuse, spam, coordinated manipulation
124 * Content can be flagged by community
125 * Audit trail maintained even if content hidden
126
127 **Federation**:
128 * Multiple nodes with independent governance
129 * No single point of control
130 * Cross-node contradiction detection
131 * Trust model prevents malicious node influence
132
133 ----
134
135 == What happens when new evidence contradicts an existing verdict? ==
136
137 FactHarbor is designed for evolving knowledge:
138
139 **Automatic Re-evaluation**:
140 1. New evidence arrives
141 2. System detects affected scenarios and verdicts
142 3. AKEL proposes updated verdicts
143 4. Reviewers/experts validate
144 5. New verdict version published
145 6. Old versions remain accessible
146
147 **Version History**:
148 * Every verdict has complete history
149 * Users can see "as of date X, what did we know?"
150 * Timeline shows how understanding evolved
151
152 **Transparent Updates**:
153 * Reason for re-evaluation documented
154 * New evidence clearly linked
155 * Changes explained, not hidden
156
157 **User Notifications**:
158 * Users following claims are notified of updates
159 * Can compare old vs new verdicts
160 * Can see which evidence changed conclusions
161
162 ----
163
164 == Who can submit claims to FactHarbor? ==
165
166 **Anyone** - even without login:
167
168 **Readers** (no login required):
169 * Browse and search all published content
170 * Submit text for analysis
171 * New claims added automatically unless duplicates exist
172 * System deduplicates and normalizes
173
174 **Contributors** (logged in):
175 * Everything Readers can do
176 * Submit evidence sources
177 * Suggest scenarios
178 * Participate in discussions
179
180 **Workflow**:
181 1. User submits text (as Reader or Contributor)
182 2. AKEL extracts claims
183 3. Checks for existing duplicates
184 4. Normalizes claim text
185 5. Assigns risk tier
186 6. Generates scenarios (draft)
187 7. Runs quality gates
188 8. Publishes as AI-Generated (Mode 2) if passes
189
190 ----
191
192 == What are "risk tiers" and why do they matter? ==
193
194 Risk tiers determine review requirements and publication workflow:
195
196 **Tier A (High Risk)**:
197 * **Domains**: Medical, legal, elections, safety, security, major financial
198 * **Publication**: AI can publish with warnings, expert review required for "Human-Reviewed" status
199 * **Audit rate**: Recommendation 30-50%
200 * **Why**: Potential for significant harm if wrong
201
202 **Tier B (Medium Risk)**:
203 * **Domains**: Complex policy, science causality, contested issues
204 * **Publication**: AI can publish immediately with clear labeling
205 * **Audit rate**: Recommendation 10-20%
206 * **Why**: Nuanced but lower immediate harm risk
207
208 **Tier C (Low Risk)**:
209 * **Domains**: Definitions, established facts, historical data
210 * **Publication**: AI publication default
211 * **Audit rate**: Recommendation 5-10%
212 * **Why**: Well-established, low controversy
213
214 **Assignment**:
215 * AKEL suggests tier based on domain, keywords, impact
216 * Moderators and Experts can override
217 * Risk tiers reviewed based on audit outcomes
218
219 ----
220
221 == How does federation work and why is it important? ==
222
223 **Federation Model**:
224 * Multiple independent FactHarbor nodes
225 * Each node has own database, AKEL, governance
226 * Nodes exchange claims, scenarios, evidence, verdicts
227 * No central authority
228
229 **Why Federation Matters**:
230 * **Resilience**: No single point of failure or censorship
231 * **Autonomy**: Communities govern themselves
232 * **Scalability**: Add nodes to handle more users
233 * **Specialization**: Domain-focused nodes (health, energy, etc.)
234 * **Trust diversity**: Multiple perspectives, not single truth source
235
236 **How Nodes Exchange Data**:
237 1. Local node creates versions
238 2. Builds signed bundle
239 3. Pushes to trusted neighbor nodes
240 4. Remote nodes validate signatures and lineage
241 5. Accept or branch versions
242 6. Local re-evaluation if needed
243
244 **Trust Model**:
245 * Trusted nodes → auto-import
246 * Neutral nodes → import with review
247 * Untrusted nodes → manual only
248
249 ----
250
251 == Can experts disagree in FactHarbor? ==
252
253 **Yes - and that's a feature, not a bug**:
254
255 **Multiple Scenarios**:
256 * Experts can create different scenarios with different assumptions
257 * Each scenario gets its own verdict
258 * Users see *why* experts disagree (different definitions, boundaries, evidence weighting)
259
260 **Parallel Verdicts**:
261 * Same scenario, different expert interpretations
262 * Both verdicts visible with expert attribution
263 * No forced consensus
264
265 **Transparency**:
266 * Expert reasoning documented
267 * Assumptions stated explicitly
268 * Evidence chains traceable
269 * Users can evaluate competing expert opinions
270
271 **Federation**:
272 * Different nodes can have different expert conclusions
273 * Cross-node branching allowed
274 * Users can see how conclusions vary across nodes
275
276 ----
277
278 == What prevents AI from hallucinating or making up facts? ==
279
280 **Multiple Safeguards**:
281
282 **Quality Gate 4: Structural Integrity**:
283 * Fact-checking against sources
284 * No hallucinations allowed
285 * Logic chain must be valid and traceable
286 * References must be accessible and verifiable
287
288 **Evidence Requirements**:
289 * Primary sources required
290 * Citations must be complete
291 * Sources must be accessible
292 * Reliability scored
293
294 **Audit System**:
295 * Human auditors check AI-generated content
296 * Hallucinations caught and fed back into training
297 * Patterns of errors trigger system improvements
298
299 **Transparency**:
300 * All reasoning chains visible
301 * Sources linked
302 * Users can verify claims against sources
303 * AKEL outputs clearly labeled
304
305 **Human Oversight**:
306 * Tier A requires expert review for "Human-Reviewed" status
307 * Audit sampling catches errors
308 * Community can flag issues
309
310 ----
311
312 == How does FactHarbor make money / is it sustainable? ==
313
314 [ToDo: Business model and sustainability to be defined]
315
316 Potential models under consideration:
317 * Non-profit foundation with grants and donations
318 * Institutional subscriptions (universities, research organizations, media)
319 * API access for third-party integrations
320 * Premium features for power users
321 * Federated node hosting services
322
323 Core principle: **Public benefit** mission takes priority over profit.
324
325 ----
326
327 == Related Pages ==
328
329 * [[Requirements (Roles)>>FactHarbor.Specification.Requirements.WebHome]]
330 * [[AKEL (AI Knowledge Extraction Layer)>>FactHarbor.Specification.AI Knowledge Extraction Layer (AKEL).WebHome]]
331 * [[Automation>>FactHarbor.Specification.Automation.WebHome]]
332 * [[Federation & Decentralization>>FactHarbor.Specification.Federation & Decentralization.WebHome]]
333 * [[Mission & Purpose>>FactHarbor.Organisation.Mission & Purpose.WebHome]]