Skip to Content

Wiki source code of Workflows

Last modified by Robert Schaub on 2026/02/08 08:25
Show last authors
1 = Workflows =
2
3 **Version:** 0.9.70
4 **Last Updated:** December 21, 2025
5 **Status:** CORRECTED - Automation Philosophy Consistent
6
7 This page describes FactHarbor's core workflows with the automation-first philosophy.
8
9
10 == 1. Core Workflow Principles ==
11
12 * **Automation First:** 90%+ content published automatically
13 * **No Approval Bottlenecks:** No centralized review queues
14 * **Quality Gates:** Automated validation before publication
15 * **Sampling Audits:** Pattern analysis for system improvement
16 * **Transparent Confidence:** All outputs labeled with confidence scores
17
18 == 2. Claim Submission Workflow ==
19
20 === 2.1 Claim Extraction ===
21
22 When users submit content (text, articles, web pages), FactHarbor first extracts individual verifiable claims:
23
24 **Input Types:**
25
26 * Single claim: "The Earth is flat"
27 * Text with multiple claims: "Climate change is accelerating. Sea levels rose 3mm in 2023. Arctic ice decreased 13% annually."
28 * URLs: Web pages analyzed for factual claims
29
30 **Extraction Process:**
31
32 * LLM analyzes submitted content
33 * Identifies distinct, verifiable factual claims
34 * Separates claims from opinions, questions, or commentary
35 * Each claim becomes independent for processing
36
37 **Output:**
38
39 * List of claims with context
40 * Each claim assigned unique ID
41 * Original context preserved for reference
42
43 This extraction ensures:
44
45 * Each claim receives focused analysis
46 * Multiple claims in one submission are all processed
47 * Claims are properly isolated for independent verification
48 * Context is preserved for accurate interpretation
49
50 **Flow:**
51 ```
52 User submits → Duplicate detection → Categorization → Processing queue → User receives ID
53 ```
54
55 **Timeline:** Seconds
56 **No approval needed:** Instant processing
57
58
59 == 3. Automated Analysis Workflow ==
60
61 **Complete Pipeline:**
62
63 ```
64 Claim from queue
65 ↓
66 Evidence gathering (AKEL)
67 ↓
68 Source evaluation (track record check)
69 ↓
70 Scenario generation
71 ↓
72 Verdict synthesis
73 ↓
74 Risk assessment
75 ↓
76 Quality gates validation
77 ↓
78 Decision: PUBLISH or BLOCK
79 ```
80
81 **Timeline:** 10-30 seconds
82 **Automation Rate:** 90%+ published automatically
83
84
85 === 3.1 Quality Gates Decision ===
86
87 **Gate Validation:**
88
89 1. Gate 1: Source Quality ✓
90 2. Gate 2: Contradiction Search ✓
91 3. Gate 3: Uncertainty Quantification ✓
92 4. Gate 4: Structural Integrity ✓
93
94 **If ALL gates PASS:**
95 → **Publish immediately** (Mode 2: AI-Generated)
96 → Apply appropriate risk tier label
97 → Display confidence score
98 → Make available for sampling audit
99
100 **If ANY gate FAILS:**
101 → **Block publication** (Mode 1: Draft-Only)
102 → Log failure reason
103 → Analyze failure pattern
104 → Queue system improvement task
105 → May re-process after improvements
106
107 **CRITICAL:** No human approval step - gates are automated.
108
109
110 == 4. Publication Workflow ==
111
112 **V0.9.70 CLARIFIED:** Risk tiers affect LABELS and AUDIT FREQUENCY, NOT approval requirements.
113
114 === Standard Flow (90%+) ===
115
116 ```
117 Pass quality gates
118 ↓
119 Determine risk tier (A/B/C)
120 ↓
121 Apply appropriate labels
122 ↓
123 PUBLISH IMMEDIATELY
124 ↓
125 Add to audit sampling pool
126 ```
127
128 **No delays, no approval queues**
129
130
131 === High-Risk Content (Tier A - <10%) ===
132
133 **V0.9.70 CORRECTION:**
134
135 ```
136 Pass quality gates
137 ↓
138 Identified as Tier A (medical/legal/safety)
139 ↓
140 PUBLISH IMMEDIATELY with prominent warnings
141 ↓
142 Higher sampling audit frequency (50%)
143 ```
144
145 **What changed from V0.9.69:**
146 - ❌ REMOVED: "Risk > 80% → Moderator review"
147 - ✅ ADDED: "Risk > 80% → Publish with WARNING labels"
148
149 **Philosophy:** Publish with strong warnings, monitor closely through sampling.
150
151 **Warning Labels for Tier A:**
152 ```
153 ⚠️ HIGH-IMPACT TOPIC
154 AI-Generated Analysis
155
156 This claim involves [medical/legal/financial/safety] topics.
157 - Confidence: [X]%
158 - Last Updated: [timestamp]
159 - This is NOT professional advice
160 - Consult qualified professionals for decisions
161
162 [View Evidence] [See Methodology] [Report Issue]
163 ```
164
165
166 === Low Quality Content (<10%) ===
167
168 ```
169 FAIL quality gates
170 ↓
171 Confidence < threshold OR structural issues
172 ↓
173 BLOCK (Mode 1: Draft-Only)
174 ↓
175 Log failure patterns
176 ↓
177 Queue for system improvement
178 ```
179
180 **NOT:** Send for human review
181 **IS:** Improve prompts/algorithms based on failure patterns
182
183
184 == 5. User Contribution Workflow ==
185
186 **Philosophy:** Wikipedia-style immediate application + audit trail
187
188 ```
189 Contributor edits published content
190 ↓
191 System validates (basic checks)
192 ↓
193 Applied IMMEDIATELY
194 ↓
195 Logged in version history
196 ↓
197 Reputation earned
198 ↓
199 May be selected for sampling audit
200 ```
201
202 **No approval required:** Changes apply instantly
203 **Quality control:** Through sampling audits and reputation system
204
205 **New contributors** (<50 reputation): Limited to minor edits
206
207
208 == 6. Sampling Audit Workflow ==
209
210 **Purpose:** Improve system quality through pattern analysis
211
212 === 6.1 Selection Process ===
213
214 ```
215 Published content
216 ↓
217 Stratified sampling (by risk tier, confidence, traffic)
218 ↓
219 Selected for audit (Tier A: 50%, B: 20%, C: 5%)
220 ↓
221 Added to audit queue
222 ```
223
224
225 === 6.2 Audit Execution ===
226
227 ```
228 Auditor receives sample
229 ↓
230 Reviews against quality standards
231 ↓
232 Identifies issues/patterns
233 ↓
234 Logs findings
235 ↓
236 System improvement tasks created
237 ```
238
239 **What auditors DO:**
240
241 * ✅ Analyze patterns across multiple outputs
242 * ✅ Identify systematic issues
243 * ✅ Recommend algorithm/prompt improvements
244 * ✅ Track accuracy trends
245
246 **What auditors DON'T DO:**
247
248 * ❌ Approve individual outputs before publication
249 * ❌ Manually fix individual outputs
250 * ❌ Act as gatekeepers
251 * ❌ Override quality gates
252
253 === 6.3 Improvement Loop ===
254
255 ```
256 Audit findings aggregated
257 ↓
258 Patterns identified
259 ↓
260 System improvements proposed
261 ↓
262 Implemented and tested
263 ↓
264 Deployed
265 ↓
266 Metrics monitored
267 ```
268
269 **Examples of Improvements:**
270
271 * Refine evidence search queries
272 * Adjust source reliability weights
273 * Enhance contradiction detection
274 * Improve claim extraction prompts
275 * Recalibrate risk tier thresholds
276
277 == 7. Flagging Workflow ==
278
279 **Two types of flags:**
280
281 === 7.1 Quality Issues ===
282
283 ```
284 User flags quality issue
285 ↓
286 Categorized automatically
287 ↓
288 Added to sampling audit pool (priority)
289 ↓
290 Pattern analysis
291 ↓
292 System improvement if pattern found
293 ```
294
295 **NOT:** Manual correction of individual claim
296 **IS:** Improve system to prevent similar issues
297
298
299 === 7.2 Abuse/Spam ===
300
301 ```
302 User flags abuse/spam
303 ↓
304 Automated pre-moderation check
305 ↓
306 Moderator review (if needed)
307 ↓
308 Action taken (hide/ban)
309 ```
310
311 **Moderator role:** Handle abuse/spam, NOT content quality
312
313
314 == 8. Moderation Workflow ==
315
316 **V0.9.70 CLARIFIED:** Moderators handle ABUSE, not content quality
317
318 === 8.1 Content Moderation (Abuse/Spam) ===
319
320 **Moderator Queue Contains:**
321
322 * Flagged abusive content
323 * Spam detection alerts
324 * Harassment reports
325 * Privacy violations
326 * Terms of service violations
327
328 **Moderator Actions:**
329
330 * Hide abusive content
331 * Ban repeat offenders
332 * Handle appeals
333 * Escalate to governing team
334
335 **Moderators DO NOT:**
336
337 * ❌ Approve content for publication
338 * ❌ Review content quality before publication
339 * ❌ Act as editorial gatekeepers
340 * ❌ Manually fix AI outputs
341
342 === 8.2 Appeal Process ===
343
344 ```
345 User disagrees with moderation
346 ↓
347 Appeals to different moderator
348 ↓
349 If still disagrees, escalates to Governing Team
350 ↓
351 Governing Team decision (final)
352 ```
353
354
355 == 9. Time Evolution Workflow ==
356
357 **Automatic Re-evaluation:**
358
359 ```
360 Published claim
361 ↓
362 Monitoring for triggers:
363 - New evidence published
364 - Source retractions
365 - Significant events
366 - Scheduled review (6-12 months)
367 ↓
368 Trigger detected
369 ↓
370 AKEL re-processes claim
371 ↓
372 Quality gates validate
373 ↓
374 If verdict changes: Correction workflow
375 ↓
376 If passes: Update published analysis
377 ```
378
379 **Correction Workflow (New in V0.9.70):**
380
381 ```
382 Verdict changed significantly
383 ↓
384 Generate correction notice
385 ↓
386 Publish correction banner (30 days)
387 ↓
388 Update corrections log
389 ↓
390 Notify users (email, RSS, API)
391 ↓
392 Update ClaimReview schema
393 ```
394
395
396 == 10. Contributor Journey ==
397
398 1. **Visitor** – Explores platform, reads documentation
399 2. **New Contributor** – Submits first improvements (typo fixes, clarifications)
400 3. **Contributor** – Contributes regularly, follows conventions
401 4. **Trusted Contributor** – Track record of quality work
402 5. **Reviewer** – Participates in sampling audits (pattern analysis)
403 6. **Moderator** – Handles abuse/spam (not content quality)
404 7. **Expert** (optional) – Provides domain expertise for contested claims
405
406 **All contributions apply immediately** - no approval workflow
407
408
409 == 11. Related Pages ==
410
411 * [[AKEL>>Archive.FactHarbor 2026\.01\.20.Specification.AI Knowledge Extraction Layer (AKEL).WebHome]] - AI processing system
412 * [[Architecture>>Archive.FactHarbor 2026\.01\.20.Specification.Architecture.WebHome]] - System architecture
413 * [[Requirements>>Archive.FactHarbor 2026\.01\.20.Specification.Requirements.WebHome]] - Requirements and roles
414 * [[Decision Processes>>FactHarbor.Organisation.Decision-Processes.WebHome]] - Governance
415
416 **V0.9.70 CHANGES:**
417
418 **REMOVED:**
419 - ❌ "High Risk → Moderator review" (was approval workflow)
420 - ❌ "Review queue" language for publication
421 - ❌ Any implication that moderators approve content quality
422
423 **ADDED/CLARIFIED:**
424 - ✅ Risk tiers affect warnings and audit frequency, NOT approval
425 - ✅ High-risk content publishes immediately with prominent warnings
426 - ✅ Quality gate failures → Block + improve system (not human review)
427 - ✅ Clear distinction: Sampling audits (improvement) vs. Content moderation (abuse)
428 - ✅ Moderator role clarified: Abuse only, NOT content quality
429 - ✅ User contributions apply immediately (Wikipedia model)
430 - ✅ Correction workflow for significant verdict changes
431 - ✅ Time evolution and re-evaluation workflow