Wiki source code of Workflows

Version 4.1 by Robert Schaub on 2025/12/12 15:41

Show last authors
1 = Workflows =
2
3 This chapter defines the core workflows used across the FactHarbor system.
4
5 Each workflow describes:
6
7 * Purpose
8 * Participants
9 * Steps
10 * Automation vs. manual work
11
12 Workflows included:
13
14 1. Claim Workflow
15 2. Scenario Workflow
16 3. Evidence Workflow
17 4. Verdict Workflow
18 5. Re-evaluation Workflow
19 6. Federation Synchronization Workflow
20 7. User Role & Review Workflow
21 8. AKEL Workflow
22 9. Global Trigger Flow
23 10. Entity Lifecycle Notes
24
25 ----
26
27 == Claim Workflow ==
28
29 **Purpose:**
30 Transform raw text or input material into a normalized, classified, deduplicated, and versioned claim ready for scenario evaluation.
31
32 **Participants:**
33 * Contributor
34 * AKEL
35 * Reviewer
36
37 === Steps ===
38
39 1. **Ingestion**
40 * User submits text, URL, transcript, or multi-claim content
41 * AKEL extracts one or multiple claims
42
43 1. **Normalization**
44 * Standardizes wording
45 * Reduces ambiguity
46 * Flags implicit assumptions
47
48 1. **Classification (AKEL draft → Reviewer confirm)**
49 * ClaimType
50 * Domain
51 * Evaluability
52 * SafetyCategory
53
54 1. **Duplicate & Similarity Detection**
55 * Embeddings created
56 * Similar claims found
57 * Reviewer merges, splits, or confirms uniqueness
58
59 1. **Version Creation**
60 * New ClaimVersion stored
61 * Every edit creates a new immutable version
62
63 1. **Cluster Assignment**
64 * AKEL proposes cluster membership
65 * Reviewer confirms
66
67 1. **Scenario Linking (optional)**
68 * Existing scenarios connected
69 * AKEL may propose new drafts
70
71 1. **Publication**
72 * Claim becomes active and visible
73
74 **Flow:**
75 Ingest → Normalize → Classify → Deduplicate → Cluster → Version → Publish
76
77 ----
78
79 == Scenario Workflow ==
80
81 **Purpose:**
82 Define the specific analytic contexts needed to evaluate each claim.
83
84 **Participants:**
85 * Contributor
86 * Reviewer
87 * Domain Expert
88 * AKEL
89
90 === Steps ===
91
92 1. **Scenario Proposal**
93 * Drafted by contributor or generated by AKEL
94
95 1. **Completion of Required Fields**
96 Must include:
97 * Definitions
98 * Assumptions
99 * ContextBoundary
100 * EvaluationMethod
101 * SafetyClass
102 * VersionMetadata
103
104 1. **Safety Interception (AKEL)**
105 Flags:
106 * non-falsifiable structures
107 * pseudoscientific assumptions
108 * unsafe contexts
109
110 1. **Redundancy & Conflict Check**
111 * Similar scenarios merged
112 * Contradictions flagged
113
114 1. **Reviewer Validation**
115 Ensures clarity, neutrality, and methodological validity.
116
117 1. **Expert Approval (mandatory for high-risk domains)**
118
119 1. **Version Storage**
120 * Each revision = new ScenarioVersion
121
122 **Flow:**
123 Draft → Validate → Safety Check → Review → Expert Approval → Version → Activate
124
125 ----
126
127 == Evidence Workflow ==
128
129 **Purpose:**
130 Structure, classify, validate, version, and link evidence to scenarios.
131
132 **Participants:**
133 * Contributor
134 * Reviewer
135 * Domain Expert
136 * AKEL
137
138 === Steps ===
139
140 1. **Evidence Submission**
141 * File, dataset, URL, or extracted text
142
143 1. **Metadata Extraction (AKEL)**
144 * EvidenceType
145 * Category
146 * Provenance
147 * Study design
148 * ExtractionMethod
149 * ReliabilityHints
150
151 1. **Relevance Check**
152 Reviewer verifies which scenarios the evidence applies to.
153
154 1. **Reliability Assessment**
155 * AKEL proposes reliability
156 * Reviewer confirms
157 * Expert review for complex papers
158
159 1. **ScenarioEvidenceLink Creation**
160 Each link stores:
161 * relevance score
162 * justification
163 * evidence version
164
165 1. **Versioning**
166 * Any update = new EvidenceVersion
167
168 **Flow:**
169 Submit → Extract Metadata → Evaluate Relevance → Score Reliability → Link → Version
170
171 ----
172
173 == Verdict Workflow ==
174
175 **Purpose:**
176 Generate likelihood estimates per scenario based on evidence and scenario structure.
177
178 **Participants:**
179 * AKEL (drafts)
180 * Reviewer
181 * Domain Expert
182
183 === Steps ===
184
185 1. **Evidence Aggregation**
186 Collect relevant evidence versions.
187
188 1. **Draft Verdict Generation (AKEL)**
189 Outputs:
190 * likelihood range
191 * uncertainty factors
192 * conflict detection
193 * sensitivity analysis
194
195 1. **Reasoning Draft**
196 Structured explanation chain generated by AKEL.
197
198 1. **Reviewer Validation**
199 Ensures logic, evidence fit, no hallucinations.
200
201 1. **Expert Review**
202 Required for:
203 * medicine
204 * psychology
205 * engineering
206 * political misinformation
207 * controversial or risky domains
208
209 1. **Verdict Storage**
210 * Every update creates a new VerdictVersion
211
212 **Flow:**
213 Aggregate → Draft Verdict → Draft Explanation → Review → Expert Approval → Version
214
215 ----
216
217 == Re-evaluation Workflow ==
218
219 **Purpose:**
220 Keep verdicts current when evidence or scenarios change.
221
222 === Trigger Types ===
223
224 * Evidence updated, disputed, retracted
225 * Scenario assumptions changed
226 * Claim reclassification
227 * AKEL contradiction detection
228 * Federation sync
229
230 === Steps ===
231
232 1. **Trigger Detection**
233 Re-evaluation engine receives event.
234
235 1. **Impact Analysis**
236 Find affected:
237 * scenarios
238 * evidence links
239 * verdicts
240
241 1. **AKEL Draft Re-calculation**
242 New:
243 * likelihood
244 * reasoning
245 * uncertainty
246
247 1. **Reviewer Validation**
248 1. **Expert Review** (high-risk)
249 1. **Version Storage**
250
251 **Flow:**
252 Trigger → Analyze → Recompute → Review → Expert → Version
253
254 ----
255
256 == Federation Synchronization Workflow ==
257
258 **Purpose:**
259 Exchange structured data between nodes.
260
261 === Steps ===
262 1. Detect version changes
263 1. Build bundle (diff + Merkle tree + signatures)
264 1. Push to peers
265 1. Validate lineage + hashes
266 1. Resolve conflicts (merge or branch)
267 1. Optional re-evaluation
268
269 **Flow:**
270 Change → Bundle → Push → Validate → Merge/Fork → Update
271
272 ----
273
274 == User Role & Review Workflow ==
275
276 **Purpose:**
277 Ensure correctness, neutrality, safety, and resistance to manipulation.
278
279 === Steps ===
280
281 1. **Submission**
282 Claim / scenario / evidence / verdict.
283
284 1. **Auto-check (AKEL)**
285 Flags unsafe content, contradictions, format issues.
286
287 1. **Reviewer Validation**
288
289 1. **Expert Validation**
290 Required for sensitive domains.
291
292 1. **Moderator Oversight**
293 Triggered by suspicious behavior.
294
295 **Flow:**
296 Submit → Auto-check → Review → Expert → Moderator (if needed)
297
298 ----
299
300 == AKEL Workflow ==
301
302 **Purpose:**
303 Support extraction, drafting, structuring, and contradiction detection.
304
305 === Stages ===
306
307 **A — Input Understanding:**
308 Extraction, normalization, classification.
309
310 **B — Scenario Drafting:**
311 Definitions, boundaries, assumptions.
312
313 **C — Evidence Processing:**
314 Retrieval, summarization, ranking.
315
316 **D — Verdict Drafting:**
317 Likelihood, explanations, uncertainties.
318
319 **E — Safety & Integrity:**
320 Contradictions, hallucination detection.
321
322 **F — Human Approval:**
323 Reviewer and/or expert.
324
325 **Flow:**
326 Input → Drafts → Integrity → Human Approval
327
328 ----
329
330 == Global Trigger Flow (Cascade) ==
331
332 Trigger Sources:
333 * Claim change
334 * Scenario change
335 * Evidence change
336 * Verdict contradiction
337 * Federation update
338 * AKEL model improvements
339
340 **Cascade Flow:**
341 Trigger → Dependency Graph → Re-evaluation → Updated Verdicts