Wiki source code of Requirements
Version 6.1 by Robert Schaub on 2025/12/14 18:59
Show last authors
| author | version | line-number | content |
|---|---|---|---|
| 1 | = Requirements = | ||
| 2 | |||
| 3 | This page defines **Roles**, **Responsibilities**, and **Rules** for contributors and users of FactHarbor. | ||
| 4 | |||
| 5 | == Roles == | ||
| 6 | |||
| 7 | === Contributor === | ||
| 8 | |||
| 9 | **Who**: Anyone (logged in or anonymous). | ||
| 10 | |||
| 11 | **Can**: | ||
| 12 | * Submit claims | ||
| 13 | * Submit evidence | ||
| 14 | * Provide feedback | ||
| 15 | * Suggest scenarios | ||
| 16 | * Flag content for review | ||
| 17 | * Request human review of AI-generated content | ||
| 18 | |||
| 19 | **Cannot**: | ||
| 20 | * Publish or mark content as "reviewed" or "approved" | ||
| 21 | * Override expert or maintainer decisions | ||
| 22 | * Directly modify AKEL or quality gate configurations | ||
| 23 | |||
| 24 | === Reviewer === | ||
| 25 | |||
| 26 | **Who**: Trusted community members, appointed by maintainers. | ||
| 27 | |||
| 28 | **Can**: | ||
| 29 | * Review contributions from Contributors and AKEL drafts | ||
| 30 | * Validate AI-generated content (Mode 2 → Mode 3 transition) | ||
| 31 | * Edit claims, scenarios, and evidence | ||
| 32 | * Add clarifications or warnings | ||
| 33 | * Change content status: `draft` → `in review` → `published` / `rejected` | ||
| 34 | * Approve or reject **Tier B and C** content for "Human-Reviewed" status | ||
| 35 | * Flag content for expert review | ||
| 36 | * Participate in audit sampling | ||
| 37 | |||
| 38 | **Cannot**: | ||
| 39 | * Approve Tier A content for "Human-Reviewed" status (requires Expert) | ||
| 40 | * Change governance rules | ||
| 41 | * Unilaterally change expert conclusions without process | ||
| 42 | * Bypass quality gates | ||
| 43 | |||
| 44 | **Note on AI-Drafted Content**: | ||
| 45 | * Reviewers can validate AI-generated content (Mode 2) to promote it to "Human-Reviewed" (Mode 3) | ||
| 46 | * For Tier B and C, Reviewers have approval authority | ||
| 47 | * For Tier A, only Experts can grant "Human-Reviewed" status | ||
| 48 | |||
| 49 | === Expert (Domain-Specific) === | ||
| 50 | |||
| 51 | **Who**: Subject-matter specialists in specific domains (medicine, law, science, etc.). | ||
| 52 | |||
| 53 | **Can**: | ||
| 54 | * Everything a Reviewer can do | ||
| 55 | * **Final authority** on Tier A content "Human-Reviewed" status | ||
| 56 | * Validate complex or controversial claims in their domain | ||
| 57 | * Define domain-specific quality standards | ||
| 58 | * Set reliability thresholds for domain sources | ||
| 59 | * Participate in risk tier assignment review | ||
| 60 | * Override AKEL suggestions in their domain (with documentation) | ||
| 61 | |||
| 62 | **Cannot**: | ||
| 63 | * Change platform governance policies | ||
| 64 | * Approve content outside their expertise domain | ||
| 65 | * Bypass technical quality gates (but can flag for adjustment) | ||
| 66 | |||
| 67 | **Specialization**: | ||
| 68 | * Experts are domain-specific (e.g., "Medical Expert", "Legal Expert", "Climate Science Expert") | ||
| 69 | * Cross-domain claims may require multiple expert reviews | ||
| 70 | |||
| 71 | === Auditor === | ||
| 72 | |||
| 73 | **Who**: Reviewers or Experts assigned to sampling audit duties. | ||
| 74 | |||
| 75 | **Can**: | ||
| 76 | * Review sampled AI-generated content against quality standards | ||
| 77 | * Validate quality gate enforcement | ||
| 78 | * Identify patterns in AI errors or hallucinations | ||
| 79 | * Provide feedback for system improvement | ||
| 80 | * Flag content for immediate review if errors found | ||
| 81 | * Contribute to audit statistics and transparency reports | ||
| 82 | |||
| 83 | **Cannot**: | ||
| 84 | * Change audit sampling algorithms (maintainer responsibility) | ||
| 85 | * Bypass normal review workflows | ||
| 86 | * Audit content they personally created | ||
| 87 | |||
| 88 | **Selection**: | ||
| 89 | * Auditors selected based on domain expertise and review quality | ||
| 90 | * Rotation to prevent audit fatigue | ||
| 91 | * Stratified assignment (Tier A auditors need higher expertise) | ||
| 92 | |||
| 93 | **Audit Focus**: | ||
| 94 | * Tier A: Recommendation 30-50% sampling rate, expert auditors | ||
| 95 | * Tier B: Recommendation 10-20% sampling rate, reviewer/expert auditors | ||
| 96 | * Tier C: Recommendation 5-10% sampling rate, reviewer auditors | ||
| 97 | |||
| 98 | === Moderator === | ||
| 99 | |||
| 100 | **Who**: Maintainers or trusted long-term contributors. | ||
| 101 | |||
| 102 | **Can**: | ||
| 103 | * All Reviewer and Expert capabilities (cross-domain) | ||
| 104 | * Manage user accounts and permissions | ||
| 105 | * Handle disputes and conflicts | ||
| 106 | * Enforce community guidelines | ||
| 107 | * Suspend or ban abusive users | ||
| 108 | * Finalize publication status for sensitive content | ||
| 109 | * Review and adjust risk tier assignments | ||
| 110 | * Oversee audit system performance | ||
| 111 | |||
| 112 | **Cannot**: | ||
| 113 | * Change core data model or architecture | ||
| 114 | * Override technical system constraints | ||
| 115 | * Make unilateral governance decisions without consensus | ||
| 116 | |||
| 117 | === Maintainer === | ||
| 118 | |||
| 119 | **Who**: Core team members responsible for the platform. | ||
| 120 | |||
| 121 | **Can**: | ||
| 122 | * All Moderator capabilities | ||
| 123 | * Change data model, architecture, and technical systems | ||
| 124 | * Configure quality gates and AKEL parameters | ||
| 125 | * Adjust audit sampling algorithms | ||
| 126 | * Set and modify risk tier policies | ||
| 127 | * Make platform-wide governance decisions | ||
| 128 | * Access and modify backend systems | ||
| 129 | * Deploy updates and fixes | ||
| 130 | * Grant and revoke roles | ||
| 131 | |||
| 132 | **Governance**: | ||
| 133 | * Maintainers operate under organizational governance rules | ||
| 134 | * Major policy changes require Governing Team approval | ||
| 135 | * Technical decisions made collaboratively | ||
| 136 | |||
| 137 | ---- | ||
| 138 | |||
| 139 | == Content Publication States == | ||
| 140 | |||
| 141 | === Mode 1: Draft === | ||
| 142 | * Not visible to public | ||
| 143 | * Visible to contributor and reviewers | ||
| 144 | * Can be edited by contributor or reviewers | ||
| 145 | * Default state for failed quality gates | ||
| 146 | |||
| 147 | === Mode 2: AI-Generated (Published) === | ||
| 148 | * **Public** and visible to all users | ||
| 149 | * Clearly labeled as "AI-Generated, Awaiting Human Review" | ||
| 150 | * Passed all automated quality gates | ||
| 151 | * Risk tier displayed (A/B/C) | ||
| 152 | * Users can: | ||
| 153 | ** Read and use content | ||
| 154 | ** Request human review | ||
| 155 | ** Flag for expert attention | ||
| 156 | * Subject to sampling audits | ||
| 157 | * Can be promoted to Mode 3 by reviewer/expert validation | ||
| 158 | |||
| 159 | === Mode 3: Human-Reviewed (Published) === | ||
| 160 | * **Public** and visible to all users | ||
| 161 | * Labeled as "Human-Reviewed" with reviewer/expert attribution | ||
| 162 | * Passed quality gates + human validation | ||
| 163 | * Highest trust level | ||
| 164 | * For Tier A, requires Expert approval | ||
| 165 | * For Tier B/C, Reviewer approval sufficient | ||
| 166 | |||
| 167 | === Rejected === | ||
| 168 | * Not visible to public | ||
| 169 | * Visible to contributor with rejection reason | ||
| 170 | * Can be resubmitted after addressing issues | ||
| 171 | * Rejection logged for transparency | ||
| 172 | |||
| 173 | ---- | ||
| 174 | |||
| 175 | == Contribution Rules == | ||
| 176 | |||
| 177 | === All Contributors Must: === | ||
| 178 | * Provide sources for claims | ||
| 179 | * Use clear, neutral language | ||
| 180 | * Avoid personal attacks or insults | ||
| 181 | * Respect intellectual property (cite sources) | ||
| 182 | * Accept community feedback gracefully | ||
| 183 | |||
| 184 | === AKEL (AI) Must: === | ||
| 185 | * Mark all outputs with `AuthorType = AI` | ||
| 186 | * Pass quality gates before Mode 2 publication | ||
| 187 | * Perform mandatory contradiction search | ||
| 188 | * Disclose confidence levels and uncertainty | ||
| 189 | * Provide traceable reasoning chains | ||
| 190 | * Flag potential bubbles or echo chambers | ||
| 191 | * Submit to audit sampling | ||
| 192 | |||
| 193 | === Reviewers Must: === | ||
| 194 | * Be impartial and evidence-based | ||
| 195 | * Document reasoning for decisions | ||
| 196 | * Escalate to experts when appropriate | ||
| 197 | * Participate in audits when assigned | ||
| 198 | * Provide constructive feedback | ||
| 199 | |||
| 200 | === Experts Must: === | ||
| 201 | * Stay within domain expertise | ||
| 202 | * Disclose conflicts of interest | ||
| 203 | * Document specialized terminology | ||
| 204 | * Provide reasoning for domain-specific decisions | ||
| 205 | * Participate in Tier A audits | ||
| 206 | |||
| 207 | ---- | ||
| 208 | |||
| 209 | == Quality Standards == | ||
| 210 | |||
| 211 | === Source Requirements === | ||
| 212 | * Primary sources preferred over secondary | ||
| 213 | * Publication date and author must be identifiable | ||
| 214 | * Sources must be accessible (not paywalled when possible) | ||
| 215 | * Contradictory sources must be acknowledged | ||
| 216 | * Echo chamber sources must be flagged | ||
| 217 | |||
| 218 | === Claim Requirements === | ||
| 219 | * Falsifiable or evaluable | ||
| 220 | * Clear definitions of key terms | ||
| 221 | * Boundaries and scope stated | ||
| 222 | * Assumptions made explicit | ||
| 223 | * Uncertainty acknowledged | ||
| 224 | |||
| 225 | === Evidence Requirements === | ||
| 226 | * Relevant to the claim and scenario | ||
| 227 | * Reliability assessment provided | ||
| 228 | * Methodology described (for studies) | ||
| 229 | * Limitations noted | ||
| 230 | * Conflicting evidence acknowledged | ||
| 231 | |||
| 232 | ---- | ||
| 233 | |||
| 234 | == Risk Tier Assignment == | ||
| 235 | |||
| 236 | **Automated (AKEL)**: Initial tier suggested based on domain, keywords, impact | ||
| 237 | **Human Validation**: Moderators or Experts can override AKEL suggestions | ||
| 238 | **Review**: Risk tiers periodically reviewed based on audit outcomes | ||
| 239 | |||
| 240 | **Tier A Indicators**: | ||
| 241 | * Medical diagnosis or treatment advice | ||
| 242 | * Legal interpretation or advice | ||
| 243 | * Election or voting information | ||
| 244 | * Safety or security sensitive | ||
| 245 | * Major financial decisions | ||
| 246 | * Potential for significant harm | ||
| 247 | |||
| 248 | **Tier B Indicators**: | ||
| 249 | * Complex scientific causality | ||
| 250 | * Contested policy domains | ||
| 251 | * Historical interpretation with political implications | ||
| 252 | * Significant economic impact claims | ||
| 253 | |||
| 254 | **Tier C Indicators**: | ||
| 255 | * Established historical facts | ||
| 256 | * Simple definitions | ||
| 257 | * Well-documented scientific consensus | ||
| 258 | * Basic reference information | ||
| 259 | |||
| 260 | ---- | ||
| 261 | |||
| 262 | == Related Pages == | ||
| 263 | |||
| 264 | * [[AKEL (AI Knowledge Extraction Layer)>>FactHarbor.Specification.AI Knowledge Extraction Layer (AKEL).WebHome]] | ||
| 265 | * [[Automation>>FactHarbor.Specification.Automation.WebHome]] | ||
| 266 | * [[Workflows>>FactHarbor.Specification.Workflows.WebHome]] | ||
| 267 | * [[Governance>>FactHarbor.Organisation.Governance]] |