Skip to Content

Wiki source code of Workflows

Last modified by Robert Schaub on 2025/12/22 14:16
Show last authors
1 = Workflows = **Version:** 0.9.70 **Last Updated:** December 21, 2025 **Status:** CORRECTED - Automation Philosophy Consistent This page describes FactHarbor's core workflows with the automation-first philosophy. == 1. Core Workflow Principles == * **Automation First:** 90%+ content published automatically
2 * **No Approval Bottlenecks:** No centralized review queues
3 * **Quality Gates:** Automated validation before publication
4 * **Sampling Audits:** Pattern analysis for system improvement
5 * **Transparent Confidence:** All outputs labeled with confidence scores == 2. Claim Submission Workflow == === 2.1 Claim Extraction === When users submit content (text, articles, web pages), FactHarbor first extracts individual verifiable claims: **Input Types:**
6 * Single claim: "The Earth is flat"
7 * Text with multiple claims: "Climate change is accelerating. Sea levels rose 3mm in 2023. Arctic ice decreased 13% annually."
8 * URLs: Web pages analyzed for factual claims **Extraction Process:**
9 * LLM analyzes submitted content
10 * Identifies distinct, verifiable factual claims
11 * Separates claims from opinions, questions, or commentary
12 * Each claim becomes independent for processing **Output:**
13 * List of claims with context
14 * Each claim assigned unique ID
15 * Original context preserved for reference This extraction ensures:
16 * Each claim receives focused analysis
17 * Multiple claims in one submission are all processed
18 * Claims are properly isolated for independent verification
19 * Context is preserved for accurate interpretation **Flow:**
20 ```
21 User submits → Duplicate detection → Categorization → Processing queue → User receives ID
22 ``` **Timeline:** Seconds **No approval needed:** Instant processing == 3. Automated Analysis Workflow == **Complete Pipeline:** ```
23 Claim from queue
24 ↓
25 Evidence gathering (AKEL)
26 ↓
27 Source evaluation (track record check)
28 ↓
29 Scenario generation
30 ↓
31 Verdict synthesis
32 ↓
33 Risk assessment
34 ↓
35 Quality gates validation
36 ↓
37 Decision: PUBLISH or BLOCK
38 ``` **Timeline:** 10-30 seconds **Automation Rate:** 90%+ published automatically === 3.1 Quality Gates Decision == **Gate Validation:**
39 1. Gate 1: Source Quality ✓
40 2. Gate 2: Contradiction Search ✓
41 3. Gate 3: Uncertainty Quantification ✓
42 4. Gate 4: Structural Integrity ✓ **If ALL gates PASS:**
43 → **Publish immediately** (Mode 2: AI-Generated)
44 → Apply appropriate risk tier label
45 → Display confidence score
46 → Make available for sampling audit **If ANY gate FAILS:**
47 → **Block publication** (Mode 1: Draft-Only)
48 → Log failure reason
49 → Analyze failure pattern
50 → Queue system improvement task
51 → May re-process after improvements **CRITICAL:** No human approval step - gates are automated. == 4. Publication Workflow == **V0.9.70 CLARIFIED:** Risk tiers affect LABELS and AUDIT FREQUENCY, NOT approval requirements. === Standard Flow (90%+) === ```
52 Pass quality gates
53 ↓
54 Determine risk tier (A/B/C)
55 ↓
56 Apply appropriate labels
57 ↓
58 PUBLISH IMMEDIATELY
59 ↓
60 Add to audit sampling pool
61 ``` **No delays, no approval queues** === High-Risk Content (Tier A - <10%) === **V0.9.70 CORRECTION:** ```
62 Pass quality gates
63 ↓
64 Identified as Tier A (medical/legal/safety)
65 ↓
66 PUBLISH IMMEDIATELY with prominent warnings
67 ↓
68 Higher sampling audit frequency (50%)
69 ``` **What changed from V0.9.69:**
70 - ❌ REMOVED: "Risk > 80% → Moderator review"
71 - ✅ ADDED: "Risk > 80% → Publish with WARNING labels" **Philosophy:** Publish with strong warnings, monitor closely through sampling. **Warning Labels for Tier A:**
72 ```
73 ⚠️ HIGH-IMPACT TOPIC
74 AI-Generated Analysis This claim involves [medical/legal/financial/safety] topics.
75 - Confidence: [X]%
76 - Last Updated: [timestamp]
77 - This is NOT professional advice
78 - Consult qualified professionals for decisions [View Evidence] [See Methodology] [Report Issue]
79 ``` === Low Quality Content (<10%) === ```
80 FAIL quality gates
81 ↓
82 Confidence < threshold OR structural issues
83 ↓
84 BLOCK (Mode 1: Draft-Only)
85 ↓
86 Log failure patterns
87 ↓
88 Queue for system improvement
89 ``` **NOT:** Send for human review **IS:** Improve prompts/algorithms based on failure patterns == 5. User Contribution Workflow == **Philosophy:** Wikipedia-style immediate application + audit trail ```
90 Contributor edits published content
91 ↓
92 System validates (basic checks)
93 ↓
94 Applied IMMEDIATELY
95 ↓
96 Logged in version history
97 ↓
98 Reputation earned
99 ↓
100 May be selected for sampling audit
101 ``` **No approval required:** Changes apply instantly **Quality control:** Through sampling audits and reputation system **New contributors** (<50 reputation): Limited to minor edits == 6. Sampling Audit Workflow == **Purpose:** Improve system quality through pattern analysis === 6.1 Selection Process === ```
102 Published content
103 ↓
104 Stratified sampling (by risk tier, confidence, traffic)
105 ↓
106 Selected for audit (Tier A: 50%, B: 20%, C: 5%)
107 ↓
108 Added to audit queue
109 ``` === 6.2 Audit Execution === ```
110 Auditor receives sample
111 ↓
112 Reviews against quality standards
113 ↓
114 Identifies issues/patterns
115 ↓
116 Logs findings
117 ↓
118 System improvement tasks created
119 ``` **What auditors DO:**
120 * ✅ Analyze patterns across multiple outputs
121 * ✅ Identify systematic issues
122 * ✅ Recommend algorithm/prompt improvements
123 * ✅ Track accuracy trends **What auditors DON'T DO:**
124 * ❌ Approve individual outputs before publication
125 * ❌ Manually fix individual outputs
126 * ❌ Act as gatekeepers
127 * ❌ Override quality gates === 6.3 Improvement Loop === ```
128 Audit findings aggregated
129 ↓
130 Patterns identified
131 ↓
132 System improvements proposed
133 ↓
134 Implemented and tested
135 ↓
136 Deployed
137 ↓
138 Metrics monitored
139 ``` **Examples of Improvements:**
140 * Refine evidence search queries
141 * Adjust source reliability weights
142 * Enhance contradiction detection
143 * Improve claim extraction prompts
144 * Recalibrate risk tier thresholds == 7. Flagging Workflow == **Two types of flags:** === 7.1 Quality Issues === ```
145 User flags quality issue
146 ↓
147 Categorized automatically
148 ↓
149 Added to sampling audit pool (priority)
150 ↓
151 Pattern analysis
152 ↓
153 System improvement if pattern found
154 ``` **NOT:** Manual correction of individual claim **IS:** Improve system to prevent similar issues === 7.2 Abuse/Spam === ```
155 User flags abuse/spam
156 ↓
157 Automated pre-moderation check
158 ↓
159 Moderator review (if needed)
160 ↓
161 Action taken (hide/ban)
162 ``` **Moderator role:** Handle abuse/spam, NOT content quality == 8. Moderation Workflow == **V0.9.70 CLARIFIED:** Moderators handle ABUSE, not content quality === 8.1 Content Moderation (Abuse/Spam) === **Moderator Queue Contains:**
163 * Flagged abusive content
164 * Spam detection alerts
165 * Harassment reports
166 * Privacy violations
167 * Terms of service violations **Moderator Actions:**
168 * Hide abusive content
169 * Ban repeat offenders
170 * Handle appeals
171 * Escalate to governing team **Moderators DO NOT:**
172 * ❌ Approve content for publication
173 * ❌ Review content quality before publication
174 * ❌ Act as editorial gatekeepers
175 * ❌ Manually fix AI outputs === 8.2 Appeal Process === ```
176 User disagrees with moderation
177 ↓
178 Appeals to different moderator
179 ↓
180 If still disagrees, escalates to Governing Team
181 ↓
182 Governing Team decision (final)
183 ``` == 9. Time Evolution Workflow == **Automatic Re-evaluation:** ```
184 Published claim
185 ↓
186 Monitoring for triggers: - New evidence published - Source retractions - Significant events - Scheduled review ↓
187 Trigger detected
188 ↓
189 AKEL re-processes claim
190 ↓
191 Quality gates validate
192 ↓
193 If verdict changes: Correction workflow
194 ↓
195 If passes: Update published analysis
196 ``` **Correction Workflow (New in V0.9.70):** ```
197 Verdict changed significantly
198 ↓
199 Generate correction notice
200 ↓
201 Publish correction banner (30 days)
202 ↓
203 Update corrections log
204 ↓
205 Notify users (email, RSS, API)
206 ↓
207 Update ClaimReview schema
208 ``` == 10. Contributor Journey == 1. **Visitor** – Explores platform, reads documentation
209 2. **New Contributor** – Submits first improvements (typo fixes, clarifications)
210 3. **Contributor** – Contributes regularly, follows conventions
211 4. **Trusted Contributor** – Track record of quality work
212 5. **Reviewer** – Participates in sampling audits (pattern analysis)
213 6. **Moderator** – Handles abuse/spam (not content quality)
214 7. **Expert** (optional) – Provides domain expertise for contested claims **All contributions apply immediately** - no approval workflow == 11. Related Pages == * [[AKEL>>FactHarbor.Specification.AI Knowledge Extraction Layer (AKEL).WebHome]] - AI processing system
215 * [[Architecture>>FactHarbor.Specification.Architecture.WebHome]] - System architecture
216 * [[Requirements>>FactHarbor.Specification.Requirements.WebHome]] - Requirements and roles
217 * [[Decision Processes>>FactHarbor.Organisation.Decision-Processes.WebHome]] - Governance **V0.9.70 CHANGES:** **REMOVED:**
218 - ❌ "High Risk → Moderator review" (was approval workflow)
219 - ❌ "Review queue" language for publication
220 - ❌ Any implication that moderators approve content quality **ADDED/CLARIFIED:**
221 - ✅ Risk tiers affect warnings and audit frequency, NOT approval
222 - ✅ High-risk content publishes immediately with prominent warnings
223 - ✅ Quality gate failures → Block + improve system (not human review)
224 - ✅ Clear distinction: Sampling audits (improvement) vs. Content moderation (abuse)
225 - ✅ Moderator role clarified: Abuse only, NOT content quality
226 - ✅ User contributions apply immediately (Wikipedia model)
227 - ✅ Correction workflow for significant verdict changes
228 - ✅ Time evolution and re-evaluation workflow