Skip to Content

Changes for page Workflows

Last modified by Robert Schaub on 2025/12/24 20:35
From version 1.1
edited by Robert Schaub
on 2025/12/16 20:34
Change comment: Imported from XAR
To version 1.2
edited by Robert Schaub
on 2025/12/16 20:39
Change comment: Renamed back-links.

Summary

Details

Page properties
Content
... ... @@ -20,6 +20,7 @@
20 20  **Actor**: Contributor or AKEL
21 21  
22 22  **Actions**:
23 +
23 23  * Submit claim text
24 24  * Provide initial sources (optional for human contributors, mandatory for AKEL)
25 25  * System assigns initial AuthorType (Human or AI)
... ... @@ -29,6 +29,7 @@
29 29  === 2.2 Step 2: AKEL Processing ===
30 30  
31 31  **Automated Steps**:
33 +
32 32  1. Claim extraction and normalization
33 33  2. Classification (domain, type, evaluability)
34 34  3. Risk tier assignment (A/B/C suggested)
... ... @@ -42,6 +42,7 @@
42 42  === 2.3 Step 3: Quality Gate Checkpoint ===
43 43  
44 44  **Gates Evaluated**:
47 +
45 45  * Source quality
46 46  * Contradiction search completion
47 47  * Uncertainty quantification
... ... @@ -48,6 +48,7 @@
48 48  * Structural integrity
49 49  
50 50  **Outcomes**:
54 +
51 51  * **All gates pass** → Proceed to Mode 2 publication (if Tier B or C)
52 52  * **Any gate fails** → Mode 1 (Draft), flag for human review
53 53  * **Tier A** → Mode 2 with warnings + auto-escalate to expert queue
... ... @@ -55,16 +55,19 @@
55 55  === 2.4 Step 4: Publication (Risk-Tier Dependent) ===
56 56  
57 57  **Tier C (Low Risk)**:
62 +
58 58  * **Direct to Mode 2**: AI-generated, public, clearly labeled
59 59  * User can request human review
60 60  * Sampling audit applies
61 61  
62 62  **Tier B (Medium Risk)**:
68 +
63 63  * **Direct to Mode 2**: AI-generated, public, clearly labeled
64 64  * Higher audit sampling rate
65 65  * High-engagement content may auto-escalate
66 66  
67 67  **Tier A (High Risk)**:
74 +
68 68  * **Mode 2 with warnings**: AI-generated, public, prominent disclaimers
69 69  * **Auto-escalated** to expert review queue
70 70  * User warnings displayed
... ... @@ -73,6 +73,7 @@
73 73  === 2.5 Step 5: Human Review (Optional for B/C, Escalated for A) ===
74 74  
75 75  **Triggers**:
83 +
76 76  * User requests review
77 77  * Audit flags issues
78 78  * High engagement (Tier B)
... ... @@ -79,6 +79,7 @@
79 79  * Automatic (Tier A)
80 80  
81 81  **Process**:
90 +
82 82  1. Reviewer/Expert examines claim
83 83  2. Validates quality gates
84 84  3. Checks contradiction search results
... ... @@ -86,21 +86,23 @@
86 86  5. Decision: Approve, Request Changes, or Reject
87 87  
88 88  **Outcomes**:
98 +
89 89  * **Approved** → Mode 3 (Human-Reviewed)
90 90  * **Changes Requested** → Back to contributor or AKEL for revision
91 91  * **Rejected** → Rejected status with reasoning
92 92  
93 -
94 94  == 3. Scenario Creation Workflow ==
95 95  
96 96  === 3.1 Step 1: Scenario Generation ===
97 97  
98 98  **Automated (AKEL)**:
108 +
99 99  * Generate scenarios for claim
100 100  * Define boundaries, assumptions, context
101 101  * Identify evaluation methods
102 102  
103 103  **Manual (Expert/Reviewer)**:
114 +
104 104  * Create custom scenarios
105 105  * Refine AKEL-generated scenarios
106 106  * Add domain-specific nuances
... ... @@ -108,6 +108,7 @@
108 108  === 3.2 Step 2: Scenario Validation ===
109 109  
110 110  **Quality Checks**:
122 +
111 111  * Completeness (definitions, boundaries, assumptions clear)
112 112  * Relevance to claim
113 113  * Evaluability
... ... @@ -114,6 +114,7 @@
114 114  * No circular logic
115 115  
116 116  **Risk Tier Assignment**:
129 +
117 117  * Inherits from parent claim
118 118  * Can be overridden by expert if scenario increases/decreases risk
119 119  
... ... @@ -120,19 +120,21 @@
120 120  === 3.3 Step 3: Scenario Publication ===
121 121  
122 122  **Mode 2 (AI-Generated)**:
136 +
123 123  * Tier B/C scenarios can publish immediately
124 124  * Subject to sampling audits
125 125  
126 126  **Mode 1 (Draft)**:
141 +
127 127  * Tier A scenarios default to draft
128 128  * Require expert validation for Mode 2 or Mode 3
129 129  
130 -
131 131  == 4. Evidence Evaluation Workflow ==
132 132  
133 133  === 4.1 Step 1: Evidence Search & Retrieval ===
134 134  
135 135  **AKEL Actions**:
150 +
136 136  * Search academic databases, reputable media
137 137  * **Mandatory contradiction search** (counter-evidence, reservations)
138 138  * Extract metadata (author, date, publication, methodology)
... ... @@ -139,6 +139,7 @@
139 139  * Assess source reliability
140 140  
141 141  **Quality Requirements**:
157 +
142 142  * Primary sources preferred
143 143  * Diverse perspectives included
144 144  * Echo chambers flagged
... ... @@ -147,6 +147,7 @@
147 147  === 4.2 Step 2: Evidence Summarization ===
148 148  
149 149  **AKEL Generates**:
166 +
150 150  * Summary of evidence
151 151  * Relevance assessment
152 152  * Reliability score
... ... @@ -158,6 +158,7 @@
158 158  === 4.3 Step 3: Evidence Review ===
159 159  
160 160  **Reviewer/Expert Validates**:
178 +
161 161  * Accuracy of summaries
162 162  * Appropriateness of sources
163 163  * Completeness of contradiction search
... ... @@ -164,16 +164,17 @@
164 164  * Reliability assessments
165 165  
166 166  **Outcomes**:
185 +
167 167  * **Mode 2**: Evidence summaries published as AI-generated
168 168  * **Mode 3**: After human validation
169 169  * **Mode 1**: Failed quality checks or pending expert review
170 170  
171 -
172 172  == 5. Verdict Generation Workflow ==
173 173  
174 174  === 5.1 Step 1: Verdict Computation ===
175 175  
176 176  **AKEL Computes**:
195 +
177 177  * Verdict across scenarios
178 178  * Confidence scores
179 179  * Uncertainty quantification
... ... @@ -181,6 +181,7 @@
181 181  * Limitations
182 182  
183 183  **Inputs**:
203 +
184 184  * Claim text
185 185  * Scenario definitions
186 186  * Evidence assessments
... ... @@ -189,11 +189,13 @@
189 189  === 5.2 Step 2: Verdict Validation ===
190 190  
191 191  **Quality Gates**:
212 +
192 192  * All four gates apply (source, contradiction, uncertainty, structure)
193 193  * Reasoning chain must be traceable
194 194  * Assumptions must be explicit
195 195  
196 196  **Risk Tier Check**:
218 +
197 197  * Tier A: Always requires expert validation for Mode 3
198 198  * Tier B: Mode 2 allowed, audit sampling
199 199  * Tier C: Mode 2 default
... ... @@ -201,6 +201,7 @@
201 201  === 5.3 Step 3: Verdict Publication ===
202 202  
203 203  **Mode 2 (AI-Generated Verdict)**:
226 +
204 204  * Clear labeling with confidence scores
205 205  * Uncertainty disclosure
206 206  * Links to reasoning trail
... ... @@ -207,16 +207,17 @@
207 207  * User can request expert review
208 208  
209 209  **Mode 3 (Expert-Validated Verdict)**:
233 +
210 210  * Human reviewer/expert stamp
211 211  * Additional commentary (optional)
212 212  * Highest trust level
213 213  
214 -
215 215  == 6. Audit Workflow ==
216 216  
217 217  === 6.1 Step 1: Audit Sampling Selection ===
218 218  
219 219  **Stratified Sampling**:
243 +
220 220  * Risk tier priority (A > B > C)
221 221  * Low confidence scores
222 222  * High traffic content
... ... @@ -224,6 +224,7 @@
224 224  * User flags
225 225  
226 226  **Sampling Rates** (Recommendations):
251 +
227 227  * Tier A: 30-50%
228 228  * Tier B: 10-20%
229 229  * Tier C: 5-10%
... ... @@ -231,6 +231,7 @@
231 231  === 6.2 Step 2: Audit Execution ===
232 232  
233 233  **Auditor Actions**:
259 +
234 234  1. Review sampled AI-generated content
235 235  2. Validate quality gates were properly applied
236 236  3. Check contradiction search completeness
... ... @@ -238,6 +238,7 @@
238 238  5. Identify errors or hallucinations
239 239  
240 240  **Audit Outcome**:
267 +
241 241  * **Pass**: Content remains in Mode 2, logged as validated
242 242  * **Fail**: Content flagged for review, system improvement triggered
243 243  
... ... @@ -244,6 +244,7 @@
244 244  === 6.3 Step 3: Feedback Loop ===
245 245  
246 246  **System Improvements**:
274 +
247 247  * Failed audits analyzed for patterns
248 248  * AKEL parameters adjusted
249 249  * Quality gates refined
... ... @@ -250,16 +250,17 @@
250 250  * Risk tier assignments recalibrated
251 251  
252 252  **Transparency**:
281 +
253 253  * Audit statistics published periodically
254 254  * Patterns shared with community
255 255  * System improvements documented
256 256  
257 -
258 258  == 7. Mode Transition Workflow ==
259 259  
260 260  === 7.1 Mode 1 → Mode 2 ===
261 261  
262 262  **Requirements**:
291 +
263 263  * All quality gates pass
264 264  * Risk tier B or C (or A with warnings)
265 265  * Contradiction search completed
... ... @@ -269,6 +269,7 @@
269 269  === 7.2 Mode 2 → Mode 3 ===
270 270  
271 271  **Requirements**:
301 +
272 272  * Human reviewer/expert validation
273 273  * Quality standards confirmed
274 274  * For Tier A: Expert approval required
... ... @@ -279,26 +279,29 @@
279 279  === 7.3 Mode 3 → Mode 1 (Demotion) ===
280 280  
281 281  **Rare - Only if**:
312 +
282 282  * New evidence contradicts verdict
283 283  * Error discovered in reasoning
284 284  * Source retraction
285 285  
286 286  **Process**:
318 +
287 287  1. Content flagged for re-evaluation
288 288  2. Moved to draft (Mode 1)
289 289  3. Re-processed through workflow
290 290  4. Reason for demotion documented
291 291  
292 -
293 293  == 8. User Actions Across Modes ==
294 294  
295 295  === 8.1 On Mode 1 (Draft) Content ===
296 296  
297 297  **Contributors**:
329 +
298 298  * Edit their own drafts
299 299  * Submit for review
300 300  
301 301  **Reviewers/Experts**:
334 +
302 302  * View and comment
303 303  * Request changes
304 304  * Approve for Mode 2 or Mode 3
... ... @@ -306,6 +306,7 @@
306 306  === 8.2 On Mode 2 (AI-Generated) Content ===
307 307  
308 308  **All Users**:
342 +
309 309  * Read and use content
310 310  * Request human review
311 311  * Flag for expert attention
... ... @@ -312,6 +312,7 @@
312 312  * Provide feedback
313 313  
314 314  **Reviewers/Experts**:
349 +
315 315  * Validate for Mode 3 transition
316 316  * Edit and refine
317 317  * Adjust risk tier if needed
... ... @@ -319,19 +319,20 @@
319 319  === 8.3 On Mode 3 (Human-Reviewed) Content ===
320 320  
321 321  **All Users**:
357 +
322 322  * Read with highest confidence
323 323  * Still can flag if new evidence emerges
324 324  
325 325  **Reviewers/Experts**:
362 +
326 326  * Update if needed
327 327  * Trigger re-evaluation if new evidence
328 328  
329 -
330 330  == 9. Diagram References ==
331 331  
332 332  === 9.1 Claim and Scenario Lifecycle (Overview) ===
333 333  
334 -{{include reference="Test.FactHarborV09.Organisation.Diagrams.Claim and Scenario Lifecycle (Overview).WebHome"}}
370 +{{include reference="Test.FactHarbor V0\.9\.23 Lost Data.Organisation.Diagrams.Claim and Scenario Lifecycle (Overview).WebHome"}}
335 335  
336 336  === 9.2 Claim and Scenario Workflow ===
337 337  
... ... @@ -356,4 +356,4 @@
356 356  * [[Automation>>Test.FactHarborV09.Specification.Automation.WebHome]]
357 357  * [[Requirements (Roles)>>Test.FactHarborV09.Specification.Requirements.WebHome]]
358 358  * [[Governance>>Test.FactHarborV09.Organisation.Governance]]
359 -
395 +{{/include}}