Changes for page Requirements

Summary

• Page properties (1 modified, 0 added, 0 removed)

Details

Page properties

Content

...	...	@@ -1,22 +1,61 @@
1	1	= Requirements =
2		-This page defines **Roles**, **Content States**, **Rules**, and **System Principles** for FactHarbor.
	2	+
	3	+{{info}}
	4	+**Phase Assignments:** See [[Requirements Roadmap Matrix>>Archive.FactHarbor.Roadmap.Requirements-Roadmap-Matrix.WebHome]] for which requirements are implemented in which phases.
	5	+{{/info}}
	6	+
	7	+**This page defines Roles, Content States, Rules, and System Requirements for FactHarbor.**
	8	+
3	3	**Core Philosophy:** Invest in system improvement, not manual data correction. When AI makes errors, improve the algorithm and re-process automatically.
	10	+
	11	+== Navigation ==
	12	+
	13	+* **[[User Needs>>FactHarbor.Specification.Requirements.User Needs.WebHome]]** - What users need from FactHarbor (drives these requirements)
	14	+* **This page** - How we fulfill those needs through system design
	15	+
	16	+(% class="box infomessage" %)
	17	+(((
	18	+**How to read this page:**
	19	+
	20	+1. **User Needs drive Requirements**: See [[User Needs>>FactHarbor.Specification.Requirements.User Needs.WebHome]] for what users need
	21	+2. **Requirements define implementation**: This page shows how we fulfill those needs
	22	+3. **Functional Requirements (FR)**: Specific features and capabilities
	23	+4. **Non-Functional Requirements (NFR)**: Quality attributes (performance, security, etc.)
	24	+
	25	+Each requirement references which User Needs it fulfills.
	26	+)))
	27	+
4	4	== 1. Roles ==
	29	+
	30	+**Fulfills**: UN-12 (Submit claims), UN-13 (Cite verdicts), UN-14 (API access)
	31	+
5	5	FactHarbor uses three simple roles plus a reputation system.
	33	+
6	6	=== 1.1 Reader ===
	35	+
7	7	**Who**: Anyone (no login required)
	37	+
8	8	**Can**:
	39	+
9	9	* Browse and search claims
10	10	* View scenarios, evidence, verdicts, and confidence scores
11	11	* Flag issues or errors
12	12	* Use filters, search, and visualization tools
13	13	* Submit claims automatically (new claims added if not duplicates)
	45	+
14	14	**Cannot**:
	47	+
15	15	* Modify content
16	16	* Access edit history details
	50	+
	51	+**User Needs served**: UN-1 (Trust assessment), UN-2 (Claim verification), UN-3 (Article summary with FactHarbor analysis summary), UN-4 (Social media fact-checking), UN-5 (Source tracing), UN-7 (Evidence transparency), UN-8 (Understanding disagreement), UN-12 (Submit claims), UN-17 (In-article highlighting)
	52	+
17	17	=== 1.2 Contributor ===
	54	+
18	18	**Who**: Registered users (earns reputation through contributions)
	56	+
19	19	**Can**:
	58	+
20	20	* Everything a Reader can do
21	21	* Edit claims, evidence, and scenarios
22	22	* Add sources and citations
...	...	@@ -23,18 +23,28 @@
23	23	* Suggest improvements to AI-generated content
24	24	* Participate in discussions
25	25	* Earn reputation points for quality contributions
	65	+
26	26	**Reputation System**:
	67	+
27	27	* New contributors: Limited edit privileges
28	28	* Established contributors (established reputation): Full edit access
29	29	* Trusted contributors (substantial reputation): Can approve certain changes
30	30	* Reputation earned through: Accepted edits, helpful flags, quality contributions
31	31	* Reputation lost through: Reverted edits, invalid flags, abuse
	73	+
32	32	**Cannot**:
	75	+
33	33	* Delete or hide content (only moderators)
34	34	* Override moderation decisions
	78	+
	79	+**User Needs served**: UN-13 (Cite and contribute)
	80	+
35	35	=== 1.3 Moderator ===
	82	+
36	36	**Who**: Trusted community members with proven track record, appointed by governance board
	84	+
37	37	**Can**:
	86	+
38	38	* Review flagged content
39	39	* Hide harmful or abusive content
40	40	* Resolve disputes between contributors
...	...	@@ -41,34 +41,56 @@
41	41	* Issue warnings or temporary bans
42	42	* Make final decisions on content disputes
43	43	* Access full audit logs
	93	+
44	44	**Cannot**:
	95	+
45	45	* Change governance rules
46	46	* Permanently ban users without board approval
47	47	* Override technical quality gates
	99	+
48	48	**Note**: Small team (3-5 initially), supported by automated moderation tools.
	101	+
49	49	=== 1.4 Domain Trusted Contributors (Optional, Task-Specific) ===
	103	+
50	50	**Who**: Subject matter specialists invited for specific high-stakes disputes
	105	+
51	51	**Not a permanent role**: Contacted externally when needed for contested claims in their domain
	107	+
52	52	**When used**:
	109	+
53	53	* Medical claims with life/safety implications
54	54	* Legal interpretations with significant impact
55	55	* Scientific claims with high controversy
56	56	* Technical claims requiring specialized knowledge
	114	+
57	57	**Process**:
	116	+
58	58	* Moderator identifies need for expert input
59	59	* Contact expert externally (don't require them to be users)
60	60	* Trusted Contributor provides written opinion with sources
61	61	* Opinion added to claim record
62	62	* Trusted Contributor acknowledged in claim
	122	+
	123	+**User Needs served**: UN-16 (Expert validation status)
	124	+
63	63	== 2. Content States ==
	126	+
	127	+**Fulfills**: UN-1 (Trust indicators), UN-16 (Review status transparency)
	128	+
64	64	FactHarbor uses two content states. Focus is on transparency and confidence scoring, not gatekeeping.
	130	+
65	65	=== 2.1 Published ===
	132	+
66	66	**Status**: Visible to all users
	134	+
67	67	**Includes**:
	136	+
68	68	* AI-generated analyses (default state)
69	69	* User-contributed content
70	70	* Edited/improved content
	140	+
71	71	**Quality Indicators** (displayed with content):
	142	+
72	72	* **Confidence Score**: 0-100% (AI's confidence in analysis)
73	73	* **Source Quality Score**: 0-100% (based on source track record)
74	74	* **Controversy Flag**: If high dispute/edit activity
...	...	@@ -75,14 +75,23 @@
75	75	* **Completeness Score**: % of expected fields filled
76	76	* **Last Updated**: Date of most recent change
77	77	* **Edit Count**: Number of revisions
	149	+* **Review Status**: AI-generated / Human-reviewed / Expert-validated
	150	+
78	78	**Automatic Warnings**:
	152	+
79	79	* Confidence < 60%: "Low confidence - use caution"
80	80	* Source quality < 40%: "Sources may be unreliable"
81	81	* High controversy: "Disputed - multiple interpretations exist"
82	82	* Medical/Legal/Safety domain: "Seek professional advice"
	157	+
	158	+**User Needs served**: UN-1 (Trust score), UN-9 (Methodology transparency), ~~UN-15 (Evolution timeline - Deferred)~~, UN-16 (Review status)
	159	+
83	83	=== 2.2 Hidden ===
	161	+
84	84	**Status**: Not visible to regular users (only to moderators)
	163	+
85	85	**Reasons**:
	165	+
86	86	* Spam or advertising
87	87	* Personal attacks or harassment
88	88	* Illegal content
...	...	@@ -89,22 +89,32 @@
89	89	* Privacy violations
90	90	* Deliberate misinformation (verified)
91	91	* Abuse or harmful content
	172	+
92	92	**Process**:
	174	+
93	93	* Automated detection flags for moderator review
94	94	* Moderator confirms and hides
95	95	* Original author notified with reason
96	96	* Can appeal to board if disputes moderator decision
	179	+
97	97	**Note**: Content is hidden, not deleted (for audit trail)
	181	+
98	98	== 3. Contribution Rules ==
	183	+
99	99	=== 3.1 All Contributors Must ===
	185	+
100	100	* Provide sources for factual claims
101	101	* Use clear, neutral language in FactHarbor's own summaries
102	102	* Respect others and maintain civil discourse
103	103	* Accept community feedback constructively
104	104	* Focus on improving quality, not protecting ego
	191	+
105	105	=== 3.2 AKEL (AI System) ===
	193	+
106	106	**AKEL is the primary system**. Human contributions supplement and train AKEL.
	195	+
107	107	**AKEL Must**:
	197	+
108	108	* Mark all outputs as AI-generated
109	109	* Display confidence scores prominently
110	110	* Provide source citations
...	...	@@ -111,86 +111,139 @@
111	111	* Flag uncertainty clearly
112	112	* Identify contradictions in evidence
113	113	* Learn from human corrections
	204	+
114	114	**When AKEL Makes Errors**:
	206	+
115	115	1. Capture the error pattern (what, why, how common)
116	116	2. Improve the system (better prompt, model, validation)
117	117	3. Re-process affected claims automatically
118	118	4. Measure improvement (did quality increase?)
	211	+
119	119	**Human Role**: Train AKEL through corrections, not replace AKEL
	213	+
120	120	=== 3.3 Contributors Should ===
	215	+
121	121	* Improve clarity and structure
122	122	* Add missing sources
123	123	* Flag errors for system improvement
124	124	* Suggest better ways to present information
125	125	* Participate in quality discussions
	221	+
126	126	=== 3.4 Moderators Must ===
	223	+
127	127	* Be impartial
128	128	* Document moderation decisions
129	129	* Respond to appeals promptly
130	130	* Use automated tools to scale efforts
131	131	* Focus on abuse/harm, not routine quality control
	229	+
132	132	== 4. Quality Standards ==
	231	+
	232	+**Fulfills**: UN-5 (Source reliability), UN-6 (Publisher track records), UN-7 (Evidence transparency), UN-9 (Methodology transparency)
	233	+
133	133	=== 4.1 Source Requirements ===
	235	+
134	134	**Track Record Over Credentials**:
	237	+
135	135	* Sources evaluated by historical accuracy
136	136	* Correction policy matters
137	137	* Independence from conflicts of interest
138	138	* Methodology transparency
	242	+
139	139	**Source Quality Database**:
	244	+
140	140	* Automated tracking of source accuracy
141	141	* Correction frequency
142	142	* Reliability score (updated continuously)
143	143	* Users can see source track record
	249	+
144	144	**No automatic trust** for government, academia, or media - all evaluated by track record.
	251	+
	252	+**User Needs served**: UN-5 (Source provenance), UN-6 (Publisher reliability)
	253	+
145	145	=== 4.2 Claim Requirements ===
	255	+
146	146	* Clear subject and assertion
147	147	* Verifiable with available information
148	148	* Sourced (or explicitly marked as needing sources)
149	149	* Neutral language in FactHarbor summaries
150	150	* Appropriate context provided
	261	+
	262	+**User Needs served**: UN-2 (Claim extraction and verification)
	263	+
151	151	=== 4.3 Evidence Requirements ===
	265	+
152	152	* Publicly accessible (or explain why not)
153	153	* Properly cited with attribution
154	154	* Relevant to claim being evaluated
155	155	* Original source preferred over secondary
	270	+
	271	+**User Needs served**: UN-7 (Evidence transparency)
	272	+
156	156	=== 4.4 Confidence Scoring ===
	274	+
157	157	**Automated confidence calculation based on**:
	276	+
158	158	* Source quality scores
159	159	* Evidence consistency
160	160	* Contradiction detection
161	161	* Completeness of analysis
162	162	* Historical accuracy of similar claims
	282	+
163	163	**Thresholds**:
	284	+
164	164	* < 40%: Too low to publish (needs improvement)
165	165	* 40-60%: Published with "Low confidence" warning
166	166	* 60-80%: Published as standard
167	167	* 80-100%: Published as "High confidence"
	289	+
	290	+**User Needs served**: UN-1 (Trust assessment), UN-9 (Methodology transparency)
	291	+
168	168	== 5. Automated Risk Scoring ==
	293	+
	294	+**Fulfills**: UN-10 (Manipulation detection), UN-16 (Appropriate review level)
	295	+
169	169	**Replace manual risk tiers with continuous automated scoring**.
	297	+
170	170	=== 5.1 Risk Score Calculation ===
	299	+
171	171	**Factors** (weighted algorithm):
	301	+
172	172	* **Domain sensitivity**: Medical, legal, safety auto-flagged higher
173	173	* **Potential impact**: Views, citations, spread
174	174	* **Controversy level**: Flags, disputes, edit wars
175	175	* **Uncertainty**: Low confidence, contradictory evidence
176	176	* **Source reliability**: Track record of sources used
	307	+
177	177	**Score**: 0-100 (higher = more risk)
	309	+
178	178	=== 5.2 Automated Actions ===
	311	+
179	179	* **Score > 80**: Flag for moderator review before publication
180	180	* **Score 60-80**: Publish with prominent warnings
181	181	* **Score 40-60**: Publish with standard warnings
182	182	* **Score < 40**: Publish normally
	316	+
183	183	**Continuous monitoring**: Risk score recalculated as new information emerges
	318	+
	319	+**User Needs served**: UN-10 (Detect manipulation tactics), UN-16 (Review status)
	320	+
184	184	== 6. System Improvement Process ==
	322	+
185	185	**Core principle**: Fix the system, not just the data.
	324	+
186	186	=== 6.1 Error Capture ===
	326	+
187	187	**When users flag errors or make corrections**:
	328	+
188	188	1. What was wrong? (categorize)
189	189	2. What should it have been?
190	190	3. Why did the system fail? (root cause)
191	191	4. How common is this pattern?
192	192	5. Store in ErrorPattern table (improvement queue)
193		-=== 6.2 Weekly Improvement Cycle ===
	334	+
	335	+=== 6.2 Continuous Improvement Cycle ===
	336	+
194	194	1. **Review**: Analyze top error patterns
195	195	2. **Develop**: Create fix (prompt, model, validation)
196	196	3. **Test**: Validate fix on sample claims
...	...	@@ -197,8 +197,11 @@
197	197	4. **Deploy**: Roll out if quality improves
198	198	5. **Re-process**: Automatically update affected claims
199	199	6. **Monitor**: Track quality metrics
	343	+
200	200	=== 6.3 Quality Metrics Dashboard ===
	345	+
201	201	**Track continuously**:
	347	+
202	202	* Error rate by category
203	203	* Source quality distribution
204	204	* Confidence score trends
...	...	@@ -206,158 +206,1753 @@
206	206	* Correction acceptance rate
207	207	* Re-work rate
208	208	* Claims processed per hour
209		-**Goal**: 10% monthly improvement in error rate
	355	+
	356	+**Goal**: continuous improvement in error rate
	357	+
210	210	== 7. Automated Quality Monitoring ==
	359	+
211	211	**Replace manual audit sampling with automated monitoring**.
	361	+
212	212	=== 7.1 Continuous Metrics ===
	363	+
213	213	* **Source quality**: Track record database
214	214	* **Consistency**: Contradiction detection
215	215	* **Clarity**: Readability scores
216	216	* **Completeness**: Field validation
217	217	* **Accuracy**: User corrections tracked
	369	+
218	218	=== 7.2 Anomaly Detection ===
	371	+
219	219	**Automated alerts for**:
	373	+
220	220	* Sudden quality drops
221	221	* Unusual patterns
222	222	* Contradiction clusters
223	223	* Source reliability changes
224	224	* User behavior anomalies
	379	+
225	225	=== 7.3 Targeted Review ===
	381	+
226	226	* Review only flagged items
227	227	* Random sampling for calibration (not quotas)
228	228	* Learn from corrections to improve automation
229		-== 8. Claim Intake & Normalization ==
230		-=== 8.1 FR1 – Claim Intake ===
	385	+
	386	+== 8. Functional Requirements ==
	387	+
	388	+This section defines specific features that fulfill user needs.
	389	+
	390	+=== 8.1 Claim Intake & Normalization ===
	391	+
	392	+==== FR1 — Claim Intake ====
	393	+
	394	+**Fulfills**: UN-2 (Claim extraction), UN-4 (Quick fact-checking), UN-12 (Submit claims)
	395	+
231	231	* Users submit claims via simple form or API
232	232	* Claims can be text, URL, or image
233	233	* Duplicate detection (semantic similarity)
234	234	* Auto-categorization by domain
235		-=== 8.2 FR2 – Claim Normalization ===
	400	+
	401	+==== FR2 — Claim Normalization ====
	402	+
	403	+**Fulfills**: UN-2 (Claim verification)
	404	+
236	236	* Standardize to clear assertion format
237	237	* Extract key entities (who, what, when, where)
238	238	* Identify claim type (factual, predictive, evaluative)
239	239	* Link to existing similar claims
240		-=== 8.3 FR3 – Claim Classification ===
	409	+
	410	+==== FR3 — Claim Classification ====
	411	+
	412	+**Fulfills**: UN-11 (Filtered research)
	413	+
241	241	* Domain: Politics, Science, Health, etc.
242	242	* Type: Historical fact, current stat, prediction, etc.
243	243	* Risk score: Automated calculation
244	244	* Complexity: Simple, moderate, complex
245		-== 9. Scenario System ==
246		-=== 9.1 FR4 – Scenario Generation ===
	418	+
	419	+=== 8.2 Scenario System ===
	420	+
	421	+==== FR4 — Scenario Generation ====
	422	+
	423	+**Fulfills**: UN-2 (Context-dependent verification), UN-3 (Article summary with FactHarbor analysis summary), UN-8 (Understanding disagreement)
	424	+
247	247	**Automated scenario creation**:
248		-* AKEL analyzes claim and generates likely scenarios
249		-* Each scenario includes: assumptions, evidence, conclusion
	426	+
	427	+* AKEL analyzes claim and generates likely scenarios (use-cases and contexts)
	428	+* Each scenario includes: assumptions, definitions, boundaries, evidence context
250	250	* Users can flag incorrect scenarios
251	251	* System learns from corrections
252		-=== 9.2 FR5 – Evidence Linking ===
	431	+
	432	+**Key Concept**: Scenarios represent different interpretations or contexts (e.g., "Clinical trials with healthy adults" vs. "Real-world data with diverse populations")
	433	+
	434	+==== FR5 — Evidence Linking ====
	435	+
	436	+**Fulfills**: UN-5 (Source tracing), UN-7 (Evidence transparency)
	437	+
253	253	* Automated evidence discovery from sources
254	254	* Relevance scoring
255	255	* Contradiction detection
256	256	* Source quality assessment
257		-=== 9.3 FR6 – Scenario Comparison ===
	442	+
	443	+==== FR6 — Scenario Comparison ====
	444	+
	445	+**Fulfills**: UN-3 (Article summary with FactHarbor analysis summary), UN-8 (Understanding disagreement)
	446	+
258	258	* Side-by-side comparison interface
259		-* Highlight key differences
260		-* Show evidence supporting each
261		-* Display confidence scores
262		-== 10. Verdicts & Analysis ==
263		-=== 10.1 FR7 – Automated Verdicts ===
264		-* AKEL generates verdict based on evidence
	448	+* Highlight key differences between scenarios
	449	+* Show evidence supporting each scenario
	450	+* Display confidence scores per scenario
	451	+
	452	+=== 8.3 Verdicts & Analysis ===
	453	+
	454	+==== FR7 — Automated Verdicts ====
	455	+
	456	+**Fulfills**: UN-1 (Trust score), UN-2 (Verification verdicts), UN-3 (Article summary with FactHarbor analysis summary), UN-13 (Cite verdicts)
	457	+
	458	+* AKEL generates verdict based on evidence within each scenario
	459	+* **Likelihood range** displayed (e.g., "0.70-0.85 (likely true)") - NOT binary true/false
	460	+* **Uncertainty factors** explicitly listed (e.g., "Small sample sizes", "Long-term effects unknown")
265	265	* Confidence score displayed prominently
266		-* Source quality indicators
	462	+* Source quality indicators shown
267	267	* Contradictions noted
268	268	* Uncertainty acknowledged
269		-=== 10.2 FR8 – Time Evolution ===
270		-* Claims update as new evidence emerges
271		-* Version history maintained
	465	+
	466	+**Key Innovation**: Detailed probabilistic verdicts with explicit uncertainty, not binary judgments
	467	+
	468	+==== FR8 — Time Evolution ====
	469	+
	470	+{{warning}}
	471	+**Status:** Deferred (Not in V1.0)
	472	+
	473	+This requirement has been **dropped from the current architecture and design**. Versioned entities have been replaced with simple edit history tracking only. Full evolution timeline functionality is deferred to future releases beyond V1.0.
	474	+{{/warning}}
	475	+
	476	+**Fulfills**: UN-15 (Verdict evolution timeline)
	477	+
	478	+* Claims and verdicts update as new evidence emerges
	479	+* Version history maintained for all verdicts
272	272	* Changes highlighted
273	273	* Confidence score trends visible
274		-== 11. Workflow & Moderation ==
275		-=== 11.1 FR9 – Publication Workflow ===
	482	+* Users can see "as of date X, what did we know?"
	483	+
	484	+=== 8.4 User Interface & Presentation ===
	485	+
	486	+==== FR12 — Two-Panel Summary View (Article Summary with FactHarbor Analysis Summary) ====
	487	+
	488	+**Fulfills**: UN-3 (Article Summary with FactHarbor Analysis Summary)
	489	+
	490	+**Purpose**: Provide side-by-side comparison of what a document claims vs. FactHarbor's complete analysis of its credibility
	491	+
	492	+**Left Panel: Article Summary**:
	493	+
	494	+* Document title, source, and claimed credibility
	495	+* "The Big Picture" - main thesis or position change
	496	+* "Key Findings" - structured summary of document's main claims
	497	+* "Reasoning" - document's explanation for positions
	498	+* "Conclusion" - document's bottom line
	499	+
	500	+**Right Panel: FactHarbor Analysis Summary**:
	501	+
	502	+* FactHarbor's independent source credibility assessment
	503	+* Claim-by-claim verdicts with confidence scores
	504	+* Methodology assessment (strengths, limitations)
	505	+* Overall verdict on document quality
	506	+* Analysis ID for reference
	507	+
	508	+**Design Principles**:
	509	+
	510	+* No scrolling required - both panels visible simultaneously
	511	+* Visual distinction between "what they say" and "FactHarbor's analysis"
	512	+* Color coding for verdicts (supported, uncertain, refuted)
	513	+* Confidence percentages clearly visible
	514	+* Mobile responsive (panels stack vertically on small screens)
	515	+
	516	+**Implementation Notes**:
	517	+
	518	+* Generated automatically by AKEL for every analyzed document
	519	+* Updates when verdict evolves (maintains version history)
	520	+* Exportable as standalone summary report
	521	+* Shareable via permanent URL
	522	+
	523	+==== FR13 — In-Article Claim Highlighting ====
	524	+
	525	+**Fulfills**: UN-17 (In-article claim highlighting)
	526	+
	527	+**Purpose**: Enable readers to quickly assess claim credibility while reading by visually highlighting factual claims with color-coded indicators
	528	+
	529	+==== Visual Example: Article with Highlighted Claims ====
	530	+
	531	+(% class="box" %)
	532	+(((
	533	+**Article: "New Study Shows Benefits of Mediterranean Diet"**
	534	+
	535	+A recent study published in the Journal of Nutrition has revealed new findings about the Mediterranean diet.
	536	+
	537	+(% class="box successmessage" style="margin:10px 0;" %)
	538	+(((
	539	+🟢 **Researchers found that Mediterranean diet followers had a 25% lower risk of heart disease compared to control groups**
	540	+
	541	+(% style="font-size:0.9em; color:#666;" %)
	542	+↑ WELL SUPPORTED • 87% confidence
	543	+[[Click for evidence details →]]
	544	+
	545	+
	546	+)))
	547	+
	548	+The study, which followed 10,000 participants over five years, showed significant improvements in cardiovascular health markers.
	549	+
	550	+(% class="box warningmessage" style="margin:10px 0;" %)
	551	+(((
	552	+🟡 **Some experts believe this diet can completely prevent heart attacks**
	553	+
	554	+(% style="font-size:0.9em; color:#666;" %)
	555	+↑ UNCERTAIN • 45% confidence
	556	+Overstated - evidence shows risk reduction, not prevention
	557	+[[Click for details →]]
	558	+
	559	+
	560	+)))
	561	+
	562	+Dr. Maria Rodriguez, lead researcher, recommends incorporating more olive oil, fish, and vegetables into daily meals.
	563	+
	564	+(% class="box errormessage" style="margin:10px 0;" %)
	565	+(((
	566	+🔴 **The study proves that saturated fats cause heart disease**
	567	+
	568	+(% style="font-size:0.9em; color:#666;" %)
	569	+↑ REFUTED • 15% confidence
	570	+Claim not supported by study design; correlation ≠ causation
	571	+[[Click for counter-evidence →]]
	572	+
	573	+
	574	+)))
	575	+
	576	+Participants also reported feeling more energetic and experiencing better sleep quality, though these were secondary measures.
	577	+)))
	578	+
	579	+**Legend:**
	580	+
	581	+* 🟢 = Well-supported claim (confidence ≥75%)
	582	+* 🟡 = Uncertain claim (confidence 40-74%)
	583	+* 🔴 = Refuted/unsupported claim (confidence <40%)
	584	+* Plain text = Non-factual content (context, opinions, recommendations)
	585	+
	586	+==== Tooltip on Hover/Click ====
	587	+
	588	+(% class="box infomessage" %)
	589	+(((
	590	+**FactHarbor Analysis**
	591	+
	592	+**Claim:**
	593	+"Researchers found that Mediterranean diet followers had a 25% lower risk of heart disease"
	594	+
	595	+**Verdict:** WELL SUPPORTED
	596	+**Confidence:** 87%
	597	+
	598	+**Evidence Summary:**
	599	+
	600	+* Meta-analysis of 12 RCTs confirms 23-28% risk reduction
	601	+* Consistent findings across multiple populations
	602	+* Published in peer-reviewed journal (high credibility)
	603	+
	604	+**Uncertainty Factors:**
	605	+
	606	+* Exact percentage varies by study (20-30% range)
	607	+
	608	+[[View Full Analysis →]]
	609	+)))
	610	+
	611	+**Color-Coding System**:
	612	+
	613	+* **Green**: Well-supported claims (confidence ≥75%, strong evidence)
	614	+* **Yellow/Orange**: Uncertain claims (confidence 40-74%, conflicting or limited evidence)
	615	+* **Red**: Refuted or unsupported claims (confidence <40%, contradicted by evidence)
	616	+* **Gray/Neutral**: Non-factual content (opinions, questions, procedural text)
	617	+
	618	+==== Interactive Highlighting Example (Detailed View) ====
	619	+
	620	+(% style="width:100%; border-collapse:collapse;" %)
	621	+|=**Article Text**|=**Status**|=**Analysis**
	622	+|(((
	623	+A recent study published in the Journal of Nutrition has revealed new findings about the Mediterranean diet.
	624	+)))|(% style="text-align:center;" %)Plain text|(% style="font-style:italic; color:#888;" %)Context - no highlighting
	625	+|(((
	626	+//Researchers found that Mediterranean diet followers had a 25% lower risk of heart disease compared to control groups//
	627	+)))|(% style="background-color:#D4EDDA; text-align:center; padding:8px;" %)🟢 **WELL SUPPORTED**|(((
	628	+**87% confidence**
	629	+
	630	+Meta-analysis of 12 RCTs confirms 23-28% risk reduction
	631	+
	632	+[[View Full Analysis]]
	633	+)))
	634	+|(((
	635	+The study, which followed 10,000 participants over five years, showed significant improvements in cardiovascular health markers.
	636	+)))|(% style="text-align:center;" %)Plain text|(% style="font-style:italic; color:#888;" %)Methodology - no highlighting
	637	+|(((
	638	+//Some experts believe this diet can completely prevent heart attacks//
	639	+)))|(% style="background-color:#FFF3CD; text-align:center; padding:8px;" %)🟡 **UNCERTAIN**|(((
	640	+**45% confidence**
	641	+
	642	+Overstated - evidence shows risk reduction, not prevention
	643	+
	644	+[[View Details]]
	645	+)))
	646	+|(((
	647	+Dr. Rodriguez recommends incorporating more olive oil, fish, and vegetables into daily meals.
	648	+)))|(% style="text-align:center;" %)Plain text|(% style="font-style:italic; color:#888;" %)Recommendation - no highlighting
	649	+|(((
	650	+//The study proves that saturated fats cause heart disease//
	651	+)))|(% style="background-color:#F8D7DA; text-align:center; padding:8px;" %)🔴 **REFUTED**|(((
	652	+**15% confidence**
	653	+
	654	+Claim not supported by study; correlation ≠ causation
	655	+
	656	+[[View Counter-Evidence]]
	657	+)))
	658	+
	659	+**Design Notes:**
	660	+
	661	+* Highlighted claims use italics to distinguish from plain text
	662	+* Color backgrounds match XWiki message box colors (success/warning/error)
	663	+* Status column shows verdict prominently
	664	+* Analysis column provides quick summary with link to details
	665	+
	666	+**User Actions**:
	667	+
	668	+* **Hover** over highlighted claim → Tooltip appears
	669	+* **Click** highlighted claim → Detailed analysis modal/panel
	670	+* **Toggle** button to turn highlighting on/off
	671	+* **Keyboard**: Tab through highlighted claims
	672	+
	673	+**Interaction Design**:
	674	+
	675	+* Hover/click on highlighted claim → Show tooltip with:
	676	+* Claim text
	677	+* Verdict (e.g., "WELL SUPPORTED")
	678	+* Confidence score (e.g., "85%")
	679	+* Brief evidence summary
	680	+* Link to detailed analysis
	681	+* Toggle highlighting on/off (user preference)
	682	+* Adjustable color intensity for accessibility
	683	+
	684	+**Technical Requirements**:
	685	+
	686	+* Real-time highlighting as page loads (non-blocking)
	687	+* Claim boundary detection (start/end of assertion)
	688	+* Handle nested or overlapping claims
	689	+* Preserve original article formatting
	690	+* Work with various content formats (HTML, plain text, PDFs)
	691	+
	692	+**Performance Requirements**:
	693	+
	694	+* Highlighting renders within 500ms of page load
	695	+* No perceptible delay in reading experience
	696	+* Efficient DOM manipulation (avoid reflows)
	697	+
	698	+**Accessibility**:
	699	+
	700	+* Color-blind friendly palette (use patterns/icons in addition to color)
	701	+* Screen reader compatible (ARIA labels for claim credibility)
	702	+* Keyboard navigation to highlighted claims
	703	+
	704	+**Implementation Notes**:
	705	+
	706	+* Claims extracted and analyzed by AKEL during initial processing
	707	+* Highlighting data stored as annotations with byte offsets
	708	+* Client-side rendering of highlights based on verdict data
	709	+* Mobile responsive (tap instead of hover)
	710	+
	711	+=== 8.5 Workflow & Moderation ===
	712	+
	713	+==== FR9 — Publication Workflow ====
	714	+
	715	+**Fulfills**: UN-1 (Fast access to verified content), UN-16 (Clear review status)
	716	+
276	276	**Simple flow**:
	718	+
277	277	1. Claim submitted
278	278	2. AKEL processes (automated)
279		-3. If confidence > threshold: Publish
	721	+3. If confidence > threshold: Publish (labeled as AI-generated)
280	280	4. If confidence < threshold: Flag for improvement
281	281	5. If risk score > threshold: Flag for moderator
	724	+
282	282	**No multi-stage approval process**
283		-=== 11.2 FR10 – Moderation ===
	726	+
	727	+==== FR10 — Moderation ====
	728	+
284	284	**Focus on abuse, not routine quality**:
	730	+
285	285	* Automated abuse detection
286	286	* Moderators handle flags
287	287	* Quick response to harmful content
288	288	* Minimal involvement in routine content
289		-=== 11.3 FR11 – Audit Trail ===
	735	+
	736	+==== FR11 — Audit Trail ====
	737	+
	738	+**Fulfills**: UN-14 (API access to histories), UN-15 (Evolution tracking)
	739	+
290	290	* All edits logged
291	291	* Version history public
292	292	* Moderation decisions documented
293	293	* System improvements tracked
294		-== 12. Technical Requirements ==
295		-=== 12.1 NFR1 – Performance ===
	744	+
	745	+== 9. Non-Functional Requirements ==
	746	+
	747	+=== 9.1 NFR1 — Performance ===
	748	+
	749	+**Fulfills**: UN-4 (Fast fact-checking), UN-11 (Responsive filtering)
	750	+
296	296	* Claim processing: < 30 seconds
297	297	* Search response: < 2 seconds
298	298	* Page load: < 3 seconds
299	299	* 99% uptime
300		-=== 12.2 NFR2 – Scalability ===
	755	+
	756	+=== 9.2 NFR2 — Scalability ===
	757	+
	758	+**Fulfills**: UN-14 (API access at scale)
	759	+
301	301	* Handle 10,000 claims initially
302	302	* Scale to 1M+ claims
303	303	* Support 100K+ concurrent users
304	304	* Automated processing scales linearly
305		-=== 12.3 NFR3 – Transparency ===
	764	+
	765	+=== 9.3 NFR3 — Transparency ===
	766	+
	767	+**Fulfills**: UN-7 (Evidence transparency), UN-9 (Methodology transparency), UN-13 (Citable verdicts), UN-15 (Evolution visibility)
	768	+
306	306	* All algorithms open source
307	307	* All data exportable
308	308	* All decisions documented
309	309	* Quality metrics public
310		-=== 12.4 NFR4 – Security & Privacy ===
	773	+
	774	+=== 9.4 NFR4 — Security & Privacy ===
	775	+
311	311	* Follow [[Privacy Policy>>FactHarbor.Organisation.How-We-Work-Together.Privacy-Policy]]
312	312	* Secure authentication
313	313	* Data encryption
314	314	* Regular security audits
315		-=== 12.5 NFR5 – Maintainability ===
	780	+
	781	+=== 9.5 NFR5 — Maintainability ===
	782	+
316	316	* Modular architecture
317	317	* Automated testing
318	318	* Continuous integration
319	319	* Comprehensive documentation
320		-== 13. MVP Scope ==
321		-**Phase 1 (Months 1-3): Read-Only MVP**
322		-Build:
323		-* Automated claim analysis
324		-* Confidence scoring
325		-* Source evaluation
326		-* Browse/search interface
327		-* User flagging system
328		-**Goal**: Prove AI quality before adding user editing
329		-**Phase 2 (Months 4-6): User Contributions**
330		-Add only if needed:
331		-* Simple editing (Wikipedia-style)
332		-* Reputation system
333		-* Basic moderation
334		-**Phase 3 (Months 7-12): Refinement**
335		-* Continuous quality improvement
336		-* Feature additions based on real usage
337		-* Scale infrastructure
338		-**Deferred**:
339		-* Federation (until multiple successful instances exist)
340		-* Complex contribution workflows (focus on automation)
341		-* Extensive role hierarchy (keep simple)
342		-== 14. Success Metrics ==
343		-**System Quality** (track weekly):
344		-* Error rate by category (target: -10%/month)
345		-* Average confidence score (target: increase)
346		-* Source quality distribution (target: more high-quality)
347		-* Contradiction detection rate (target: increase)
348		-**Efficiency** (track monthly):
349		-* Claims processed per hour (target: increase)
350		-* Human hours per claim (target: decrease)
351		-* Automation coverage (target: >90%)
352		-* Re-work rate (target: <5%)
353		-**User Satisfaction** (track quarterly):
354		-* User flag rate (issues found)
355		-* Correction acceptance rate (flags valid)
356		-* Return user rate
357		-* Trust indicators (surveys)
358		-== 15. Related Pages ==
359		-* [[Architecture>>FactHarbor.Specification.Architecture.WebHome]]
360		-* [[Data Model>>FactHarbor.Specification.Data Model.WebHome]]
361		-* [[Workflows>>FactHarbor.Specification.Workflows.WebHome]]
362		-* [[Global Rules>>FactHarbor.Organisation.How-We-Work-Together.GlobalRules.WebHome]]
	787	+
	788	+=== NFR11: AKEL Quality Assurance Framework ===
	789	+
	790	+**Fulfills:** AI safety, IFCN methodology transparency
	791	+
	792	+**Specification:**
	793	+
	794	+Multi-layer AI quality gates to detect hallucinations, low-confidence results, and logical inconsistencies.
	795	+
	796	+==== Quality Gate 1: Claim Extraction Validation ====
	797	+
	798	+**Purpose:** Ensure extracted claims are factual assertions (not opinions/predictions)
	799	+
	800	+**Checks:**
	801	+
	802	+1. **Factual Statement Test:** Is this verifiable? (Yes/No)
	803	+2. **Opinion Detection:** Contains hedging language? ("I think", "probably", "best")
	804	+3. **Future Prediction Test:** Makes claims about future events?
	805	+4. **Specificity Score:** Contains specific entities, numbers, dates?
	806	+
	807	+**Thresholds:**
	808	+
	809	+* Factual: Must be "Yes"
	810	+* Opinion markers: <2 hedging phrases
	811	+* Specificity: ≥3 specific elements
	812	+
	813	+**Action if Failed:** Flag as "Non-verifiable", do NOT generate verdict
	814	+
	815	+==== Quality Gate 2: Evidence Relevance Validation ====
	816	+
	817	+**Purpose:** Ensure AI-linked evidence actually relates to claim
	818	+
	819	+**Checks:**
	820	+
	821	+1. **Semantic Similarity Score:** Evidence vs. claim (embeddings)
	822	+2. **Entity Overlap:** Shared people/places/things?
	823	+3. **Topic Relevance:** Discusses claim subject?
	824	+
	825	+**Thresholds:**
	826	+
	827	+* Similarity: ≥0.6 (cosine similarity)
	828	+* Entity overlap: ≥1 shared entity
	829	+* Topic relevance: ≥0.5
	830	+
	831	+**Action if Failed:** Discard irrelevant evidence
	832	+
	833	+==== Quality Gate 3: Scenario Coherence Check ====
	834	+
	835	+**Purpose:** Validate scenario assumptions are logical and complete
	836	+
	837	+**Checks:**
	838	+
	839	+1. **Completeness:** All required fields populated
	840	+2. **Internal Consistency:** Assumptions don't contradict
	841	+3. **Distinguishability:** Scenarios meaningfully different
	842	+
	843	+**Thresholds:**
	844	+
	845	+* Required fields: 100%
	846	+* Contradiction score: <0.3
	847	+* Scenario similarity: <0.8
	848	+
	849	+**Action if Failed:** Merge duplicates, reduce confidence -20%
	850	+
	851	+==== Quality Gate 4: Verdict Confidence Assessment ====
	852	+
	853	+**Purpose:** Only publish high-confidence verdicts
	854	+
	855	+**Checks:**
	856	+
	857	+1. **Evidence Count:** Minimum 2 sources
	858	+2. **Source Quality:** Average reliability ≥0.6
	859	+3. **Evidence Agreement:** Supporting vs. contradicting ≥0.6
	860	+4. **Uncertainty Factors:** Hedging in reasoning
	861	+
	862	+**Confidence Tiers:**
	863	+
	864	+* **HIGH (80-100%):** ≥3 sources, ≥0.7 quality, ≥80% agreement
	865	+* **MEDIUM (50-79%):** ≥2 sources, ≥0.6 quality, ≥60% agreement
	866	+* **LOW (0-49%):** <2 sources OR low quality/agreement
	867	+* **INSUFFICIENT:** <2 sources → DO NOT PUBLISH
	868	+
	869	+**Implementation Phases:**
	870	+
	871	+* **POC1:** Gates 1 & 4 only (basic validation)
	872	+* **POC2:** All 4 gates (complete framework)
	873	+* **V1.0:** Hardened with <5% hallucination rate
	874	+
	875	+**Acceptance Criteria:**
	876	+
	877	+* ✅ All gates operational
	878	+* ✅ Hallucination rate <5%
	879	+* ✅ Quality metrics public
	880	+
	881	+=== NFR12: Security Controls ===
	882	+
	883	+**Fulfills:** Data protection, system integrity, user privacy, production readiness
	884	+
	885	+**Purpose:** Protect FactHarbor systems, user data, and operations from security threats, ensuring production-grade security posture.
	886	+
	887	+**Specification:**
	888	+
	889	+==== API Security ====
	890	+
	891	+**Rate Limiting:**
	892	+
	893	+* **Analysis endpoints:** 100 requests/hour per IP
	894	+* **Read endpoints:** 1,000 requests/hour per IP
	895	+* **Search:** 500 requests/hour per IP
	896	+* **Authenticated users:** 5x higher limits
	897	+* **Burst protection:** Max 10 requests/second
	898	+
	899	+**Authentication & Authorization:**
	900	+
	901	+* **API Keys:** Required for programmatic access
	902	+* **JWT tokens:** For user sessions (1-hour expiry)
	903	+* **OAuth2:** For third-party integrations
	904	+* **Role-Based Access Control (RBAC):**
	905	+* Public: Read-only access to published claims
	906	+* Contributor: Submit claims, provide evidence
	907	+* Moderator: Review contributions, manage quality
	908	+* Admin: System configuration, user management
	909	+
	910	+**CORS Policies:**
	911	+
	912	+* Whitelist approved domains only
	913	+* No wildcard origins in production
	914	+* Credentials required for sensitive endpoints
	915	+
	916	+**Input Sanitization:**
	917	+
	918	+* Validate all user input against schemas
	919	+* Sanitize HTML/JavaScript in text submissions
	920	+* Prevent SQL injection (use parameterized queries)
	921	+* Prevent command injection (no shell execution of user input)
	922	+* Max request size: 10MB
	923	+* File upload restrictions: Whitelist file types, scan for malware
	924	+
	925	+----
	926	+
	927	+==== Data Security ====
	928	+
	929	+**Encryption at Rest:**
	930	+
	931	+* Database encryption using AES-256
	932	+* Encrypted backups
	933	+* Key management via cloud provider KMS (AWS KMS, Google Cloud KMS)
	934	+* Regular key rotation (90-day cycle)
	935	+
	936	+**Encryption in Transit:**
	937	+
	938	+* HTTPS/TLS 1.3 only (no TLS 1.0/1.1)
	939	+* Strong cipher suites only
	940	+* HSTS (HTTP Strict Transport Security) enabled
	941	+* Certificate pinning for mobile apps
	942	+
	943	+**Secure Credential Storage:**
	944	+
	945	+* Passwords hashed with bcrypt (cost factor 12+)
	946	+* API keys encrypted in database
	947	+* Secrets stored in environment variables (never in code)
	948	+* Use secrets manager (AWS Secrets Manager, HashiCorp Vault)
	949	+
	950	+**Data Privacy:**
	951	+
	952	+* Minimal data collection (privacy by design)
	953	+* User data deletion on request (GDPR compliance)
	954	+* PII encryption in database
	955	+* Anonymize logs (no PII in log files)
	956	+
	957	+----
	958	+
	959	+==== Application Security ====
	960	+
	961	+**OWASP Top 10 Compliance:**
	962	+
	963	+1. **Broken Access Control:** RBAC implementation, path traversal prevention
	964	+2. **Cryptographic Failures:** Strong encryption, secure key management
	965	+3. **Injection:** Parameterized queries, input validation
	966	+4. **Insecure Design:** Security review of all features
	967	+5. **Security Misconfiguration:** Hardened defaults, security headers
	968	+6. **Vulnerable Components:** Dependency scanning (see below)
	969	+7. **Authentication Failures:** Strong password policy, MFA support
	970	+8. **Data Integrity Failures:** Signature verification, checksums
	971	+9. **Security Logging Failures:** Comprehensive audit logs
	972	+10. **Server-Side Request Forgery:** URL validation, whitelist domains
	973	+
	974	+**Security Headers:**
	975	+
	976	+* `Content-Security-Policy`: Strict CSP to prevent XSS
	977	+* `X-Frame-Options`: DENY (prevent clickjacking)
	978	+* `X-Content-Type-Options`: nosniff
	979	+* `Referrer-Policy`: strict-origin-when-cross-origin
	980	+* `Permissions-Policy`: Restrict browser features
	981	+
	982	+**Dependency Vulnerability Scanning:**
	983	+
	984	+* **Tools:** Snyk, Dependabot, npm audit, pip-audit
	985	+* **Frequency:** Daily automated scans
	986	+* **Action:** Patch critical vulnerabilities within 24 hours
	987	+* **Policy:** No known high/critical CVEs in production
	988	+
	989	+**Security Audits:**
	990	+
	991	+* **Internal:** Quarterly security reviews
	992	+* **External:** Annual penetration testing by certified firm
	993	+* **Bug Bounty:** Public bug bounty program (V1.1+)
	994	+* **Compliance:** SOC 2 Type II certification target (V1.5)
	995	+
	996	+----
	997	+
	998	+==== Operational Security ====
	999	+
	1000	+**DDoS Protection:**
	1001	+
	1002	+* CloudFlare or AWS Shield
	1003	+* Rate limiting at CDN layer
	1004	+* Automatic IP blocking for abuse patterns
	1005	+
	1006	+**Monitoring & Alerting:**
	1007	+
	1008	+* Real-time security event monitoring
	1009	+* Alerts for:
	1010	+* Failed login attempts (>5 in 10 minutes)
	1011	+* API abuse patterns
	1012	+* Unusual data access patterns
	1013	+* Security scan detections
	1014	+* Integration with SIEM (Security Information and Event Management)
	1015	+
	1016	+**Incident Response:**
	1017	+
	1018	+* Documented incident response plan
	1019	+* Security incident classification (P1-P4)
	1020	+* On-call rotation for security issues
	1021	+* Post-mortem for all security incidents
	1022	+* Public disclosure policy (coordinated disclosure)
	1023	+
	1024	+**Backup & Recovery:**
	1025	+
	1026	+* Daily encrypted backups
	1027	+* 30-day retention period
	1028	+* Tested recovery procedures (quarterly)
	1029	+* Disaster recovery plan (RTO: 4 hours, RPO: 1 hour)
	1030	+
	1031	+----
	1032	+
	1033	+==== Compliance & Standards ====
	1034	+
	1035	+**GDPR Compliance:**
	1036	+
	1037	+* User consent management
	1038	+* Right to access data
	1039	+* Right to deletion
	1040	+* Data portability
	1041	+* Privacy policy published
	1042	+
	1043	+**Accessibility:**
	1044	+
	1045	+* WCAG 2.1 AA compliance
	1046	+* Screen reader compatibility
	1047	+* Keyboard navigation
	1048	+* Alt text for images
	1049	+
	1050	+**Browser Support:**
	1051	+
	1052	+* Modern browsers only (Chrome/Edge/Firefox/Safari latest 2 versions)
	1053	+* No IE11 support
	1054	+
	1055	+**Acceptance Criteria:**
	1056	+
	1057	+* ✅ Passes OWASP ZAP security scan (no high/critical findings)
	1058	+* ✅ All dependencies with known vulnerabilities patched
	1059	+* ✅ Penetration test completed with no critical findings
	1060	+* ✅ Rate limiting blocks abuse attempts
	1061	+* ✅ Encryption at rest and in transit verified
	1062	+* ✅ Security headers scored A+ on securityheaders.com
	1063	+* ✅ Incident response plan documented and tested
	1064	+* ✅ 95% uptime over 30-day period
	1065	+
	1066	+=== NFR13: Quality Metrics Transparency ===
	1067	+
	1068	+**Fulfills:** User trust, transparency, continuous improvement, IFCN methodology transparency
	1069	+
	1070	+**Purpose:** Provide transparent, measurable quality metrics that demonstrate AKEL's performance and build user trust in automated fact-checking.
	1071	+
	1072	+**Specification:**
	1073	+
	1074	+==== Component: Public Quality Dashboard ====
	1075	+
	1076	+**Core Metrics to Display:**
	1077	+
	1078	+* \\
	1079	+** \\
	1080	+**1. Verdict Quality Metrics
	1081	+
	1082	+**TIGERScore (Fact-Checking Quality):**
	1083	+
	1084	+* **Definition:** Measures how well generated verdicts match expert fact-checker judgments
	1085	+* **Scale:** 0-100 (higher is better)
	1086	+* **Calculation:** Using TIGERScore framework (Truth-conditional accuracy, Informativeness, Generality, Evaluativeness, Relevance)
	1087	+* **Target:** Average ≥80 for production release
	1088	+* **Display:**
	1089	+{{code}}Verdict Quality (TIGERScore):
	1090	+Overall: 84.2 ▲ (+2.1 from last month)
	1091	+
	1092	+Distribution:
	1093	+ Excellent (>80): 67%
	1094	+ Good (60-80): 28%
	1095	+ Needs Improvement (<60): 5%
	1096	+
	1097	+Trend: [Graph showing improvement over time]{{/code}}
	1098	+
	1099	+**2. Hallucination & Faithfulness Metrics**
	1100	+
	1101	+**AlignScore (Faithfulness to Evidence):**
	1102	+
	1103	+* **Definition:** Measures how well verdicts align with actual evidence content
	1104	+* **Scale:** 0-1 (higher is better)
	1105	+* **Purpose:** Detect AI hallucinations (making claims not supported by evidence)
	1106	+* **Target:** Average ≥0.85, hallucination rate <5%
	1107	+* **Display:**
	1108	+{{code}}Evidence Faithfulness (AlignScore):
	1109	+Average: 0.87 ▼ (-0.02 from last month)
	1110	+
	1111	+Hallucination Rate: 4.2%
	1112	+ - Claims without evidence support: 3.1%
	1113	+ - Misrepresented evidence: 1.1%
	1114	+
	1115	+Action: Prompt engineering review scheduled{{/code}}
	1116	+
	1117	+**3. Evidence Quality Metrics**
	1118	+
	1119	+**Source Reliability:**
	1120	+
	1121	+* Average source quality score (0-1 scale)
	1122	+* Distribution of high/medium/low quality sources
	1123	+* Publisher track record trends
	1124	+
	1125	+**Evidence Coverage:**
	1126	+
	1127	+* Average number of sources per claim
	1128	+* Percentage of claims with ≥2 sources (EFCSN minimum)
	1129	+* Geographic diversity of sources
	1130	+
	1131	+**Display:**
	1132	+{{code}}Evidence Quality:
	1133	+
	1134	+Average Sources per Claim: 4.2
	1135	+Claims with ≥2 sources: 94% (EFCSN compliant)
	1136	+
	1137	+Source Quality Distribution:
	1138	+ High quality (>0.8): 48%
	1139	+ Medium quality (0.5-0.8): 43%
	1140	+ Low quality (<0.5): 9%
	1141	+
	1142	+Geographic Diversity: 23 countries represented{{/code}}
	1143	+
	1144	+**4. Contributor Consensus Metrics** (when human reviewers involved)
	1145	+
	1146	+**Inter-Rater Reliability (IRR):**
	1147	+
	1148	+* **Calculation:** Cohen's Kappa or Fleiss' Kappa for multiple raters
	1149	+* **Scale:** 0-1 (higher is better)
	1150	+* **Interpretation:**
	1151	+* >0.8: Almost perfect agreement
	1152	+* 0.6-0.8: Substantial agreement
	1153	+* 0.4-0.6: Moderate agreement
	1154	+* <0.4: Poor agreement
	1155	+* **Target:** Maintain ≥0.7 (substantial agreement)
	1156	+
	1157	+**Display:**
	1158	+{{code}}Contributor Consensus:
	1159	+
	1160	+Inter-Rater Reliability (IRR): 0.73 (Substantial agreement)
	1161	+ - Verdict agreement: 78%
	1162	+ - Evidence quality agreement: 71%
	1163	+ - Scenario structure agreement: 69%
	1164	+
	1165	+Cases requiring moderator review: 12
	1166	+Moderator override rate: 8%{{/code}}
	1167	+
	1168	+----
	1169	+
	1170	+==== Quality Dashboard Implementation ====
	1171	+
	1172	+**Dashboard Location:** `/quality-metrics`
	1173	+
	1174	+**Update Frequency:**
	1175	+
	1176	+* **POC2:** Weekly manual updates
	1177	+* **Beta 0:** Daily automated updates
	1178	+* **V1.0:** Real-time metrics (updated hourly)
	1179	+
	1180	+**Dashboard Sections:**
	1181	+
	1182	+1. **Overview:** Key metrics at a glance
	1183	+2. **Verdict Quality:** TIGERScore trends and distributions
	1184	+3. **Evidence Analysis:** Source quality and coverage
	1185	+4. **AI Performance:** Hallucination rates, AlignScore
	1186	+5. **Human Oversight:** Contributor consensus, review rates
	1187	+6. **System Health:** Processing times, error rates, uptime
	1188	+
	1189	+**Example Dashboard Layout:**
	1190	+
	1191	+{{code}}
	1192	+┌─────────────────────────────────────────────────────────────┐
	1193	+│ FactHarbor Quality Metrics Last updated: │
	1194	+│ Public Dashboard 2 hours ago │
	1195	+└─────────────────────────────────────────────────────────────┘
	1196	+
	1197	+📊 KEY METRICS
	1198	+─────────────────────────────────────────────────────────────
	1199	+TIGERScore (Verdict Quality): 84.2 ▲ (+2.1)
	1200	+AlignScore (Faithfulness): 0.87 ▼ (-0.02)
	1201	+Hallucination Rate: 4.2% ✓ (Target: <5%)
	1202	+Average Sources per Claim: 4.2 ▲ (+0.3)
	1203	+
	1204	+📈 TRENDS (30 days)
	1205	+─────────────────────────────────────────────────────────────
	1206	+[Graph: TIGERScore trending upward]
	1207	+[Graph: Hallucination rate declining]
	1208	+[Graph: Evidence quality stable]
	1209	+
	1210	+⚠️ IMPROVEMENT TARGETS
	1211	+─────────────────────────────────────────────────────────────
	1212	+1. Reduce hallucination rate to <3% (Current: 4.2%)
	1213	+2. Increase TIGERScore average to >85 (Current: 84.2)
	1214	+3. Maintain IRR >0.75 (Current: 0.73)
	1215	+
	1216	+📄 DETAILED REPORTS
	1217	+─────────────────────────────────────────────────────────────
	1218	+• Monthly Quality Report (PDF)
	1219	+• Methodology Documentation
	1220	+• AKEL Performance Analysis
	1221	+• Contributor Agreement Analysis
	1222	+
	1223	+{{/code}}
	1224	+
	1225	+----
	1226	+
	1227	+==== Continuous Improvement Feedback Loop ====
	1228	+
	1229	+**How Metrics Inform AKEL Improvements:**
	1230	+
	1231	+1. **Identify Weak Areas:**
	1232	+
	1233	+* Low TIGERScore → Review prompt engineering
	1234	+* High hallucination → Strengthen evidence grounding
	1235	+* Low IRR → Clarify evaluation criteria
	1236	+
	1237	+2. **A/B Testing Integration:**
	1238	+
	1239	+* Test prompt variations
	1240	+* Measure impact on quality metrics
	1241	+* Deploy winners automatically
	1242	+
	1243	+3. **Alert Thresholds:**
	1244	+
	1245	+* TIGERScore drops below 75 → Alert team
	1246	+* Hallucination rate exceeds 7% → Pause auto-publishing
	1247	+* IRR below 0.6 → Moderator training needed
	1248	+
	1249	+4. **Monthly Quality Reviews:**
	1250	+
	1251	+* Analyze trends
	1252	+* Identify systematic issues
	1253	+* Plan prompt improvements
	1254	+* Update AKEL models
	1255	+
	1256	+----
	1257	+
	1258	+==== Metric Calculation Details ====
	1259	+
	1260	+**TIGERScore Implementation:**
	1261	+
	1262	+* Reference: https://github.com/TIGER-AI-Lab/TIGERScore
	1263	+* Input: Generated verdict + reference verdict (from expert)
	1264	+* Output: 0-100 score across 5 dimensions
	1265	+* Requires: Test set of expert-reviewed claims (minimum 100)
	1266	+
	1267	+**AlignScore Implementation:**
	1268	+
	1269	+* Reference: https://github.com/yuh-zha/AlignScore
	1270	+* Input: Generated verdict + source evidence text
	1271	+* Output: 0-1 faithfulness score
	1272	+* Calculation: Semantic alignment between claim and evidence
	1273	+
	1274	+**Source Quality Scoring:**
	1275	+
	1276	+* Use existing source reliability database (e.g., NewsGuard, MBFC)
	1277	+* Factor in: Publication history, corrections record, transparency
	1278	+* Scale: 0-1 (weighted average across sources)
	1279	+
	1280	+----
	1281	+
	1282	+==== Integration Points ====
	1283	+
	1284	+* **NFR11: AKEL Quality Assurance** - Metrics validate quality gate effectiveness
	1285	+* **FR49: A/B Testing** - Metrics measure test success
	1286	+* **FR11: Audit Trail** - Source of quality data
	1287	+* **NFR3: Transparency** - Public metrics build trust
	1288	+
	1289	+**Acceptance Criteria:**
	1290	+
	1291	+* ✅ All core metrics implemented and calculating correctly
	1292	+* ✅ Dashboard updates daily (Beta 0) or hourly (V1.0)
	1293	+* ✅ Alerts trigger when metrics degrade beyond thresholds
	1294	+* ✅ Monthly quality report auto-generates
	1295	+* ✅ Dashboard is publicly accessible (no login required)
	1296	+* ✅ Mobile-responsive dashboard design
	1297	+* ✅ Metrics inform quarterly AKEL improvement planning
	1298	+
	1299	+== 13. Requirements Traceability ==
	1300	+
	1301	+For full traceability matrix showing which requirements fulfill which user needs, see:
	1302	+
	1303	+* [[User Needs>>FactHarbor.Specification.Requirements.User Needs.WebHome]] - Section 8 includes comprehensive mapping tables
	1304	+
	1305	+== 14. Related Pages ==
	1306	+
	1307	+**Non-Functional Requirements (see Section 9):**
	1308	+
	1309	+* [[NFR11 — AKEL Quality Assurance Framework>>#NFR11]]
	1310	+* [[NFR12 — Security Controls>>#NFR12]]
	1311	+* [[NFR13 — Quality Metrics Transparency>>#NFR13]]
	1312	+
	1313	+**Other Requirements:**
	1314	+
	1315	+* [[User Needs>>FactHarbor.Specification.Requirements.User Needs.WebHome]]
	1316	+* [[V1.0 Requirements>>FactHarbor.Specification.Requirements.V10.]]
	1317	+* [[Gap Analysis>>FactHarbor.Specification.Requirements.GapAnalysis]]
	1318	+
	1319	+* **[[User Needs>>FactHarbor.Specification.Requirements.User Needs.WebHome]]** - What users need (drives these requirements)
	1320	+* [[Architecture>>Archive.FactHarbor.Specification.Architecture.WebHome]] - How requirements are implemented
	1321	+* [[Data Model>>FactHarbor.Specification.Data Model.WebHome]] - Data structures supporting requirements
	1322	+* [[Workflows>>FactHarbor.Specification.Workflows.WebHome]] - User interaction workflows
	1323	+* [[AKEL>>Archive.FactHarbor.Specification.AI Knowledge Extraction Layer (AKEL).WebHome]] - AI system fulfilling automation requirements
	1324	+* [[Global Rules>>Archive.FactHarbor.Organisation.How-We-Work-Together.GlobalRules.WebHome]]
363	363	* [[Privacy Policy>>FactHarbor.Organisation.How-We-Work-Together.Privacy-Policy]]
	1326	+
	1327	+= V0.9.70 Additional Requirements =
	1328	+
	1329	+== Functional Requirements (Additional) ==
	1330	+
	1331	+=== FR44: ClaimReview Schema Implementation ===
	1332	+
	1333	+**Fulfills:** UN-13 (Cite FactHarbor Verdicts), UN-14 (API Access for Integration), UN-26 (Search Engine Visibility)
	1334	+
	1335	+**Purpose:** Generate valid ClaimReview structured data for every published analysis to enable Google/Bing search visibility and fact-check discovery.
	1336	+
	1337	+**Specification:**
	1338	+
	1339	+==== Component: Schema.org Markup Generator ====
	1340	+
	1341	+FactHarbor must generate valid ClaimReview structured data following Schema.org specifications for every published claim analysis.
	1342	+
	1343	+**Required JSON-LD Schema:**
	1344	+
	1345	+{{code language="json"}}
	1346	+{
	1347	+ "@context": "https://schema.org",
	1348	+ "@type": "ClaimReview",
	1349	+ "datePublished": "YYYY-MM-DD",
	1350	+ "url": "https://factharbor.org/claims/{claim_id}",
	1351	+ "claimReviewed": "The exact claim text",
	1352	+ "author": {
	1353	+ "@type": "Organization",
	1354	+ "name": "FactHarbor",
	1355	+ "url": "https://factharbor.org"
	1356	+ },
	1357	+ "reviewRating": {
	1358	+ "@type": "Rating",
	1359	+ "ratingValue": "1-5",
	1360	+ "bestRating": "5",
	1361	+ "worstRating": "1",
	1362	+ "alternateName": "FactHarbor likelihood score"
	1363	+ },
	1364	+ "itemReviewed": {
	1365	+ "@type": "Claim",
	1366	+ "author": {
	1367	+ "@type": "Person",
	1368	+ "name": "Claim author if known"
	1369	+ },
	1370	+ "datePublished": "YYYY-MM-DD if known",
	1371	+ "appearance": {
	1372	+ "@type": "CreativeWork",
	1373	+ "url": "Original claim URL if from article"
	1374	+ }
	1375	+ }
	1376	+}
	1377	+{{/code}}
	1378	+
	1379	+**FactHarbor-Specific Mapping:**
	1380	+
	1381	+**Likelihood Score to Rating Scale:**
	1382	+
	1383	+* 80-100% likelihood → 5 (Highly Supported)
	1384	+* 60-79% likelihood → 4 (Supported)
	1385	+* 40-59% likelihood → 3 (Mixed/Uncertain)
	1386	+* 20-39% likelihood → 2 (Questionable)
	1387	+* 0-19% likelihood → 1 (Refuted)
	1388	+
	1389	+**Multiple Scenarios Handling:**
	1390	+
	1391	+* If claim has multiple scenarios with different verdicts, generate **separate ClaimReview** for each scenario
	1392	+* Add `disambiguatingDescription` field explaining scenario context
	1393	+* Example: "Scenario: If interpreted as referring to 2023 data..."
	1394	+
	1395	+==== Implementation Requirements ====
	1396	+
	1397	+1. **Auto-generate** on claim publication
	1398	+2. **Embed** in HTML `<head>` section as JSON-LD script
	1399	+3. **Validate** against Schema.org validator before publishing
	1400	+4. **Submit** to Google Search Console for indexing
	1401	+5. **Update** automatically when verdict changes (integrate with FR8: Time Evolution)
	1402	+
	1403	+==== Integration Points ====
	1404	+
	1405	+* **FR7: Automated Verdicts** - Source of rating data and claim text
	1406	+* **FR8: Time Evolution** - Triggers schema updates when verdicts change
	1407	+* **FR11: Audit Trail** - Logs all schema generation and update events
	1408	+
	1409	+==== Resources ====
	1410	+
	1411	+* ClaimReview Project: https://www.claimreviewproject.com
	1412	+* Schema.org ClaimReview: https://schema.org/ClaimReview
	1413	+* Google Fact Check Guidelines: https://developers.google.com/search/docs/appearance/fact-check
	1414	+
	1415	+**Acceptance Criteria:**
	1416	+
	1417	+* ✅ Passes Google Structured Data Testing Tool
	1418	+* ✅ Appears in Google Fact Check Explorer within 48 hours of publication
	1419	+* ✅ Valid JSON-LD syntax (no errors)
	1420	+* ✅ All required fields populated with correct data types
	1421	+* ✅ Handles multi-scenario claims correctly (separate ClaimReview per scenario)
	1422	+
	1423	+=== FR45: User Corrections Notification System ===
	1424	+
	1425	+**Fulfills:** IFCN Principle 5 (Open & Honest Corrections), EFCSN compliance
	1426	+
	1427	+**Purpose:** When any claim analysis is corrected, notify users who previously viewed the claim to maintain transparency and build trust.
	1428	+
	1429	+**Specification:**
	1430	+
	1431	+==== Component: Corrections Visibility Framework ====
	1432	+
	1433	+**Correction Types:**
	1434	+
	1435	+1. **Major Correction:** Verdict changes category (e.g., "Supported" → "Refuted")
	1436	+2. **Significant Correction:** Likelihood score changes >20%
	1437	+3. **Minor Correction:** Evidence additions, source quality updates
	1438	+4. **Scenario Addition:** New scenario added to existing claim
	1439	+
	1440	+==== Notification Mechanisms ====
	1441	+
	1442	+* \\
	1443	+** \\
	1444	+**1. In-Page Banner:
	1445	+
	1446	+Display prominent banner on claim page:
	1447	+
	1448	+{{code}}
	1449	+[!] CORRECTION NOTICE
	1450	+This analysis was updated on [DATE]. [View what changed] [Dismiss]
	1451	+
	1452	+Major changes:
	1453	+• Verdict changed from "Likely True (75%)" to "Uncertain (45%)"
	1454	+• New contradicting evidence added from [Source]
	1455	+• Scenario 2 updated with additional context
	1456	+
	1457	+[See full correction log]
	1458	+{{/code}}
	1459	+
	1460	+**2. Correction Log Page:**
	1461	+
	1462	+* Public changelog at `/claims/{id}/corrections`
	1463	+* Displays for each correction:
	1464	+* Date/time of correction
	1465	+* What changed (before/after comparison)
	1466	+* Why changed (reason if provided)
	1467	+* Who made change (AKEL auto-update vs. contributor override)
	1468	+
	1469	+**3. Email Notifications (opt-in):**
	1470	+
	1471	+* Send to users who bookmarked or shared the claim
	1472	+* Subject: "FactHarbor Correction: [Claim title]"
	1473	+* Include summary of changes
	1474	+* Link to updated analysis
	1475	+
	1476	+**4. RSS/API Feed:**
	1477	+
	1478	+* Corrections feed at `/corrections.rss`
	1479	+* API endpoint: `GET /api/corrections?since={timestamp}`
	1480	+* Enables external monitoring by journalists and researchers
	1481	+
	1482	+==== Display Rules ====
	1483	+
	1484	+* Show banner on **ALL pages** displaying the claim (search results, related claims, embeddings)
	1485	+* Banner persists for **30 days** after correction
	1486	+* **"Corrections" count badge** on claim card
	1487	+* **Timestamp** on every verdict: "Last updated: [datetime]"
	1488	+
	1489	+==== IFCN Compliance Requirements ====
	1490	+
	1491	+* Corrections policy published at `/corrections-policy`
	1492	+* User can report suspected errors via `/report-error/{claim_id}`
	1493	+* Link to IFCN complaint process (if FactHarbor becomes signatory)
	1494	+* **Scrupulous transparency:** Never silently edit analyses
	1495	+
	1496	+==== Integration Points ====
	1497	+
	1498	+* **FR8: Time Evolution** - Triggers corrections when verdicts change
	1499	+* **FR11: Audit Trail** - Source of correction data and change history
	1500	+* **NFR3: Transparency** - Public correction log demonstrates commitment
	1501	+
	1502	+**Acceptance Criteria:**
	1503	+
	1504	+* ✅ Banner appears within 60 seconds of correction
	1505	+* ✅ Correction log is permanent and publicly accessible
	1506	+* ✅ Email notifications deliver within 5 minutes
	1507	+* ✅ RSS feed updates in real-time
	1508	+* ✅ Mobile-responsive banner design
	1509	+* ✅ Accessible (screen reader compatible)
	1510	+
	1511	+=== FR46: Image Verification System ===
	1512	+
	1513	+**Fulfills:** UN-27 (Visual Claim Verification)
	1514	+
	1515	+**Purpose:** Verify authenticity and context of images shared with claims to detect manipulation, misattribution, and out-of-context usage.
	1516	+
	1517	+**Specification:**
	1518	+
	1519	+==== Component: Multi-Method Image Verification ====
	1520	+
	1521	+**Method 1: Reverse Image Search**
	1522	+
	1523	+**Purpose:** Find earlier uses of the image to verify context
	1524	+
	1525	+**Implementation:**
	1526	+
	1527	+* Integrate APIs:
	1528	+* **Google Vision AI** (reverse search)
	1529	+* **TinEye** (oldest known uses)
	1530	+* **Bing Visual Search** (broad coverage)
	1531	+
	1532	+**Process:**
	1533	+
	1534	+1. Extract image from claim or user upload
	1535	+2. Query multiple reverse search services
	1536	+3. Analyze results for:
	1537	+
	1538	+* Earliest known publication
	1539	+* Original context (what was it really showing?)
	1540	+* Publication timeline
	1541	+* Geographic spread
	1542	+
	1543	+**Output:**
	1544	+{{code}}Reverse Image Search Results:
	1545	+
	1546	+Earliest known use: 2019-03-15 (5 years before claim)
	1547	+Original context: "Photo from 2019 flooding in Mumbai"
	1548	+This claim uses it for: "2024 hurricane damage in Florida"
	1549	+
	1550	+⚠️ Image is OUT OF CONTEXT
	1551	+
	1552	+Found in 47 other articles:
	1553	+• 2019-03-15: Mumbai floods (original)
	1554	+• 2020-07-22: Bangladesh monsoon
	1555	+• 2024-10-15: Current claim (misattributed)
	1556	+
	1557	+[View full timeline]{{/code}}
	1558	+
	1559	+----
	1560	+
	1561	+**Method 2: AI Manipulation Detection**
	1562	+
	1563	+**Purpose:** Detect deepfakes, face swaps, and digital alterations
	1564	+
	1565	+**Implementation:**
	1566	+
	1567	+* Integrate detection services:
	1568	+* **Sensity AI** (deepfake detection)
	1569	+* **Reality Defender** (multimodal analysis)
	1570	+* **AWS Rekognition** (face detection inconsistencies)
	1571	+
	1572	+**Detection Categories:**
	1573	+
	1574	+1. **Face Manipulation:**
	1575	+
	1576	+* Deepfake face swaps
	1577	+* Expression manipulation
	1578	+* Identity replacement
	1579	+
	1580	+2. **Image Manipulation:**
	1581	+
	1582	+* Copy-paste artifacts
	1583	+* Clone stamp detection
	1584	+* Content-aware fill detection
	1585	+* JPEG compression inconsistencies
	1586	+
	1587	+3. **AI Generation:**
	1588	+
	1589	+* Detect fully AI-generated images
	1590	+* Identify generation artifacts
	1591	+* Check for model signatures
	1592	+
	1593	+**Confidence Scoring:**
	1594	+
	1595	+* **HIGH (80-100%):** Strong evidence of manipulation
	1596	+* **MEDIUM (50-79%):** Suspicious artifacts detected
	1597	+* **LOW (0-49%):** Minor inconsistencies or inconclusive
	1598	+
	1599	+**Output:**
	1600	+{{code}}Manipulation Analysis:
	1601	+
	1602	+Face Manipulation: LOW RISK (12%)
	1603	+Image Editing: MEDIUM RISK (64%)
	1604	+ • Clone stamp artifacts detected in sky region
	1605	+ • JPEG compression inconsistent between objects
	1606	+
	1607	+AI Generation: LOW RISK (8%)
	1608	+
	1609	+⚠️ Possible manipulation detected. Manual review recommended.{{/code}}
	1610	+
	1611	+----
	1612	+
	1613	+**Method 3: Metadata Analysis (EXIF)**
	1614	+
	1615	+**Purpose:** Extract technical details that may reveal manipulation or misattribution
	1616	+
	1617	+**Extracted Data:**
	1618	+
	1619	+* **Camera/Device:** Make, model, software
	1620	+* **Timestamps:** Original date, modification dates
	1621	+* **Location:** GPS coordinates (if present)
	1622	+* **Editing History:** Software used, edit count
	1623	+* **File Properties:** Resolution, compression, format conversions
	1624	+
	1625	+**Red Flags:**
	1626	+
	1627	+* Metadata completely stripped (suspicious)
	1628	+* Timestamp conflicts with claimed date
	1629	+* GPS location conflicts with claimed location
	1630	+* Multiple edit rounds (hiding something?)
	1631	+* Creation date after modification date (impossible)
	1632	+
	1633	+**Output:**
	1634	+{{code}}Image Metadata:
	1635	+
	1636	+Camera: iPhone 14 Pro
	1637	+Original date: 2023-08-12 14:32:15
	1638	+Location: 40.7128°N, 74.0060°W (New York City)
	1639	+Modified: 2024-10-15 08:45:22
	1640	+Software: Adobe Photoshop 2024
	1641	+
	1642	+⚠️ Location conflicts with claim
	1643	+Claim says: "Taken in Los Angeles"
	1644	+EXIF says: New York City
	1645	+
	1646	+⚠️ Edited 14 months after capture{{/code}}
	1647	+
	1648	+----
	1649	+
	1650	+==== Verification Workflow ====
	1651	+
	1652	+**Automatic Triggers:**
	1653	+
	1654	+1. User submits claim with image
	1655	+2. Article being analyzed contains images
	1656	+3. Social media post includes photos
	1657	+
	1658	+**Process:**
	1659	+
	1660	+1. Extract images from content
	1661	+2. Run all 3 verification methods in parallel
	1662	+3. Aggregate results into confidence score
	1663	+4. Generate human-readable summary
	1664	+5. Display prominently in analysis
	1665	+
	1666	+**Display Integration:**
	1667	+
	1668	+Show image verification panel in claim analysis:
	1669	+
	1670	+{{code}}
	1671	+📷 IMAGE VERIFICATION
	1672	+
	1673	+[Image thumbnail]
	1674	+
	1675	+✅ Reverse Search: Original context verified
	1676	+⚠️ Manipulation: Possible editing detected (64% confidence)
	1677	+✅ Metadata: Consistent with claim details
	1678	+
	1679	+Overall Assessment: CAUTION ADVISED
	1680	+This image may have been edited. Original context appears accurate.
	1681	+
	1682	+[View detailed analysis]
	1683	+{{/code}}
	1684	+
	1685	+==== Integration Points ====
	1686	+
	1687	+* **FR7: Automated Verdicts** - Image verification affects claim credibility
	1688	+* **FR4: Analysis Summary** - Image findings included in summary
	1689	+* **UN-27: Visual Claim Verification** - Direct fulfillment
	1690	+
	1691	+==== Cost Considerations ====
	1692	+
	1693	+**API Costs (estimated per image):**
	1694	+
	1695	+* Google Vision AI: $0.001-0.003
	1696	+* TinEye: $0.02 (commercial API)
	1697	+* Sensity AI: $0.05-0.10
	1698	+* AWS Rekognition: $0.001-0.002
	1699	+
	1700	+**Total per image:** $0.07-0.15**
	1701	+
	1702	+**Mitigation Strategies:**
	1703	+
	1704	+* Cache results for duplicate images
	1705	+* Use free tier quotas where available
	1706	+* Prioritize higher-value claims for deep analysis
	1707	+* Offer premium verification as paid tier
	1708	+
	1709	+**Acceptance Criteria:**
	1710	+
	1711	+* ✅ Reverse image search finds original sources
	1712	+* ✅ Manipulation detection accuracy >80% on test dataset
	1713	+* ✅ EXIF extraction works for major image formats (JPEG, PNG, HEIC)
	1714	+* ✅ Results display within 10 seconds
	1715	+* ✅ Mobile-friendly image comparison interface
	1716	+* ✅ False positive rate <15%
	1717	+
	1718	+=== FR47: Archive.org Integration ===
	1719	+
	1720	+**Importance:** CRITICAL
	1721	+**Fulfills:** Evidence persistence, FR5 (Evidence linking)
	1722	+
	1723	+**Purpose:** Ensure evidence remains accessible even if original sources are deleted.
	1724	+
	1725	+**Specification:**
	1726	+
	1727	+**Automatic Archiving:**
	1728	+
	1729	+When AKEL links evidence:
	1730	+
	1731	+1. Check if URL already archived (Wayback Machine API)
	1732	+2. If not, submit for archiving (Save Page Now API)
	1733	+3. Store both original URL and archive URL
	1734	+4. Display both to users
	1735	+
	1736	+**Archive Display:**
	1737	+
	1738	+{{code}}
	1739	+Evidence Source: [Original URL]
	1740	+Archived: [Archive.org URL] (Captured: [date])
	1741	+
	1742	+[View Original] [View Archive]
	1743	+{{/code}}
	1744	+
	1745	+**Fallback Logic:**
	1746	+
	1747	+* If original URL unavailable → Auto-redirect to archive
	1748	+* If archive unavailable → Display warning
	1749	+* If both unavailable → Flag for manual review
	1750	+
	1751	+**API Integration:**
	1752	+
	1753	+* Use Wayback Machine Availability API
	1754	+* Use Save Page Now API (SPNv2)
	1755	+* Rate limiting: 15 requests/minute (Wayback limit)
	1756	+
	1757	+**Acceptance Criteria:**
	1758	+
	1759	+* ✅ All evidence URLs auto-archived
	1760	+* ✅ Archive links displayed to users
	1761	+* ✅ Fallback to archive if original unavailable
	1762	+* ✅ API rate limits respected
	1763	+* ✅ Archive status visible in evidence display
	1764	+
	1765	+== Category 4: Community Safety ==
	1766	+
	1767	+ FR48: Contributor Safety Framework ===
	1768	+
	1769	+**Importance:** CRITICAL
	1770	+**Fulfills:** UN-28 (Safe contribution environment)
	1771	+
	1772	+**Purpose:** Protect contributors from harassment, doxxing, and coordinated attacks.
	1773	+
	1774	+**Specification:**
	1775	+
	1776	+* \\
	1777	+** \\
	1778	+**1. Privacy Protection:
	1779	+
	1780	+* **Optional Pseudonymity:** Contributors can use pseudonyms
	1781	+* **Email Privacy:** Emails never displayed publicly
	1782	+* **Profile Privacy:** Contributors control what's public
	1783	+* **IP Logging:** Only for abuse prevention, not public
	1784	+
	1785	+**2. Harassment Prevention:**
	1786	+
	1787	+* **Automated Toxicity Detection:** Flag abusive comments
	1788	+* **Personal Information Detection:** Auto-block doxxing attempts
	1789	+* **Coordinated Attack Detection:** Identify brigading patterns
	1790	+* **Rapid Response:** Moderator alerts for harassment
	1791	+
	1792	+**3. Safety Features:**
	1793	+
	1794	+* **Block Users:** Contributors can block harassers
	1795	+* **Private Contributions:** Option to contribute anonymously
	1796	+* **Report Harassment:** One-click harassment reporting
	1797	+* **Safety Resources:** Links to support resources
	1798	+
	1799	+**4. Moderator Tools:**
	1800	+
	1801	+* **Quick Ban:** Immediately block abusers
	1802	+* **Pattern Detection:** Identify coordinated attacks
	1803	+* **Appeal Process:** Fair review of moderation actions
	1804	+* **Escalation:** Serious threats escalated to authorities
	1805	+
	1806	+**5. Trusted Contributor Protection:**
	1807	+
	1808	+* **Enhanced Privacy:** Additional protection for high-profile contributors
	1809	+* **Verification:** Optional identity verification (not public)
	1810	+* **Legal Support:** Resources for contributors facing legal threats
	1811	+
	1812	+**Acceptance Criteria:**
	1813	+
	1814	+* ✅ Pseudonyms supported
	1815	+* ✅ Toxicity detection active
	1816	+* ✅ Doxxing auto-blocked
	1817	+* ✅ Harassment reporting functional
	1818	+* ✅ Moderator tools implemented
	1819	+* ✅ Safety policy published
	1820	+
	1821	+== Category 5: Continuous Improvement ==
	1822	+
	1823	+ FR49: A/B Testing Framework ===
	1824	+
	1825	+**Importance:** CRITICAL
	1826	+**Fulfills:** Continuous system improvement
	1827	+
	1828	+**Purpose:** Test and measure improvements to AKEL prompts, algorithms, and workflows.
	1829	+
	1830	+**Specification:**
	1831	+
	1832	+**Test Capabilities:**
	1833	+
	1834	+1. **Prompt Variations:**
	1835	+
	1836	+* Test different claim extraction prompts
	1837	+* Test different verdict generation prompts
	1838	+* Measure: Accuracy, clarity, completeness
	1839	+
	1840	+2. **Algorithm Variations:**
	1841	+
	1842	+* Test different source scoring algorithms
	1843	+* Test different confidence calculations
	1844	+* Measure: Audit accuracy, user satisfaction
	1845	+
	1846	+3. **Workflow Variations:**
	1847	+
	1848	+* Test different quality gate thresholds
	1849	+* Test different risk tier assignments
	1850	+* Measure: Publication rate, quality scores
	1851	+
	1852	+**Implementation:**
	1853	+
	1854	+* **Traffic Split:** 50/50 or 90/10 splits
	1855	+* **Randomization:** Consistent per claim (not per user)
	1856	+* **Metrics Collection:** Automatic for all variants
	1857	+* **Statistical Significance:** Minimum sample size calculation
	1858	+* **Rollout:** Winner promoted to 100% traffic
	1859	+
	1860	+**A/B Test Workflow:**
	1861	+
	1862	+{{code}}
	1863	+1. Hypothesis: "New prompt improves claim extraction"
	1864	+2. Design test: Control vs. Variant
	1865	+3. Define metrics: Extraction accuracy, completeness
	1866	+4. Run test: 7-14 days, minimum 100 claims each
	1867	+5. Analyze results: Statistical significance?
	1868	+6. Decision: Deploy winner or iterate
	1869	+{{/code}}
	1870	+
	1871	+**Acceptance Criteria:**
	1872	+
	1873	+* ✅ A/B testing framework implemented
	1874	+* ✅ Can test prompt variations
	1875	+* ✅ Can test algorithm variations
	1876	+* ✅ Metrics automatically collected
	1877	+* ✅ Statistical significance calculated
	1878	+* ✅ Results inform system improvements
	1879	+
	1880	+=== FR54: Evidence Deduplication ===
	1881	+
	1882	+**Importance:** CRITICAL (POC2/Beta)
	1883	+**Fulfills:** Accurate evidence counting, quality metrics
	1884	+
	1885	+**Purpose:** Avoid counting the same source multiple times when it appears in different forms.
	1886	+
	1887	+**Specification:**
	1888	+
	1889	+**Deduplication Logic:**
	1890	+
	1891	+1. **URL Normalization:**
	1892	+
	1893	+* Remove tracking parameters (?utm_source=...)
	1894	+* Normalize http/https
	1895	+* Normalize www/non-www
	1896	+* Handle redirects
	1897	+
	1898	+2. **Content Similarity:**
	1899	+
	1900	+* If two sources have >90% text similarity → Same source
	1901	+* If one is subset of other → Same source
	1902	+* Use fuzzy matching for minor differences
	1903	+
	1904	+3. **Cross-Domain Syndication:**
	1905	+
	1906	+* Detect wire service content (AP, Reuters)
	1907	+* Mark as single source if syndicated
	1908	+* Count original publication only
	1909	+
	1910	+**Display:**
	1911	+
	1912	+{{code}}
	1913	+Evidence Sources (3 unique, 5 total):
	1914	+
	1915	+1. Original Article (NYTimes)
	1916	+ - Also appeared in: WashPost, Guardian (syndicated)
	1917	+
	1918	+2. Research Paper (Nature)
	1919	+
	1920	+3. Official Statement (WHO)
	1921	+{{/code}}
	1922	+
	1923	+**Acceptance Criteria:**
	1924	+
	1925	+* ✅ URL normalization works
	1926	+* ✅ Content similarity detected
	1927	+* ✅ Syndicated content identified
	1928	+* ✅ Unique vs. total counts accurate
	1929	+* ✅ Improves evidence quality metrics
	1930	+
	1931	+== Additional Requirements (Lower Importance) ==
	1932	+
	1933	+ FR50: OSINT Toolkit Integration ===
	1934	+
	1935	+**Fulfills:** Advanced media verification
	1936	+
	1937	+**Purpose:** Integrate open-source intelligence tools for advanced verification.
	1938	+
	1939	+**Tools to Integrate:**
	1940	+
	1941	+* InVID/WeVerify (video verification)
	1942	+* Bellingcat toolkit
	1943	+* Additional TBD based on V1.0 learnings
	1944	+
	1945	+=== FR51: Video Verification System ===
	1946	+
	1947	+**Fulfills:** UN-27 (Visual claims), advanced media verification
	1948	+
	1949	+**Purpose:** Verify video-based claims.
	1950	+
	1951	+**Specification:**
	1952	+
	1953	+* Keyframe extraction
	1954	+* Reverse video search
	1955	+* Deepfake detection (AI-powered)
	1956	+* Metadata analysis
	1957	+* Acoustic signature analysis
	1958	+
	1959	+=== FR52: Interactive Detection Training ===
	1960	+
	1961	+**Fulfills:** Media literacy education
	1962	+
	1963	+**Purpose:** Teach users to identify misinformation.
	1964	+
	1965	+**Specification:**
	1966	+
	1967	+* Interactive tutorials
	1968	+* Practice exercises
	1969	+* Detection quizzes
	1970	+* Gamification elements
	1971	+
	1972	+=== FR53: Cross-Organizational Sharing ===
	1973	+
	1974	+**Fulfills:** Collaboration with other fact-checkers
	1975	+
	1976	+**Purpose:** Share findings with IFCN/EFCSN members.
	1977	+
	1978	+**Specification:**
	1979	+
	1980	+* API for fact-checking organizations
	1981	+* Structured data exchange
	1982	+* Privacy controls
	1983	+* Attribution requirements
	1984	+
	1985	+== Summary ==
	1986	+
	1987	+**V1.0 Critical Requirements (Must Have):**
	1988	+
	1989	+* FR44: ClaimReview Schema ✅
	1990	+* FR45: Corrections Notification ✅
	1991	+* FR46: Image Verification ✅
	1992	+* FR47: Archive.org Integration ✅
	1993	+* FR48: Contributor Safety ✅
	1994	+* FR49: A/B Testing ✅
	1995	+* FR54: Evidence Deduplication ✅
	1996	+* NFR11: Quality Assurance Framework ✅
	1997	+* NFR12: Security Controls ✅
	1998	+* NFR13: Quality Metrics Dashboard ✅
	1999	+
	2000	+**V1.1+ (Future):**
	2001	+
	2002	+* FR50: OSINT Integration
	2003	+* FR51: Video Verification
	2004	+* FR52: Detection Training
	2005	+* FR53: Cross-Org Sharing
	2006	+
	2007	+**Total:** 11 critical requirements for V1.0
	2008	+
	2009	+=== FR54: Evidence Deduplication ===
	2010	+
	2011	+**Fulfills:** Accurate evidence counting, quality metrics
	2012	+
	2013	+**Purpose:** Avoid counting the same source multiple times when it appears in different forms.
	2014	+
	2015	+**Specification:**
	2016	+
	2017	+**Deduplication Logic:**
	2018	+
	2019	+1. **URL Normalization:**
	2020	+
	2021	+* Remove tracking parameters (?utm_source=...)
	2022	+* Normalize http/https
	2023	+* Normalize www/non-www
	2024	+* Handle redirects
	2025	+
	2026	+2. **Content Similarity:**
	2027	+
	2028	+* If two sources have >90% text similarity → Same source
	2029	+* If one is subset of other → Same source
	2030	+* Use fuzzy matching for minor differences
	2031	+
	2032	+3. **Cross-Domain Syndication:**
	2033	+
	2034	+* Detect wire service content (AP, Reuters)
	2035	+* Mark as single source if syndicated
	2036	+* Count original publication only
	2037	+
	2038	+**Display:**
	2039	+
	2040	+{{code}}
	2041	+Evidence Sources (3 unique, 5 total):
	2042	+
	2043	+1. Original Article (NYTimes)
	2044	+ - Also appeared in: WashPost, Guardian (syndicated)
	2045	+
	2046	+2. Research Paper (Nature)
	2047	+
	2048	+3. Official Statement (WHO)
	2049	+{{/code}}
	2050	+
	2051	+**Acceptance Criteria:**
	2052	+
	2053	+* ✅ URL normalization works
	2054	+* ✅ Content similarity detected
	2055	+* ✅ Syndicated content identified
	2056	+* ✅ Unique vs. total counts accurate
	2057	+* ✅ Improves evidence quality metrics
	2058	+
	2059	+== Additional Requirements (Lower Importance) ==
	2060	+
	2061	+ FR7: Automated Verdicts (Enhanced with Quality Gates) ===
	2062	+
	2063	+**POC1+ Enhancement:**
	2064	+
	2065	+After AKEL generates verdict, it passes through quality gates:
	2066	+
	2067	+{{code}}
	2068	+Workflow:
	2069	+1. Extract claims
	2070	+ ↓
	2071	+2. [GATE 1] Validate fact-checkable
	2072	+ ↓
	2073	+3. Generate scenarios
	2074	+ ↓
	2075	+4. Generate verdicts
	2076	+ ↓
	2077	+5. [GATE 4] Validate confidence
	2078	+ ↓
	2079	+6. Display to user
	2080	+{{/code}}
	2081	+
	2082	+**Updated Verdict States:**
	2083	+
	2084	+* PUBLISHED
	2085	+* INSUFFICIENT_EVIDENCE
	2086	+* NON_FACTUAL_CLAIM
	2087	+* PROCESSING
	2088	+* ERROR
	2089	+
	2090	+=== FR4: Analysis Summary (Enhanced with Quality Metadata) ===
	2091	+
	2092	+**POC1+ Enhancement:**
	2093	+
	2094	+Display quality indicators:
	2095	+
	2096	+{{code}}
	2097	+Analysis Summary:
	2098	+ Verifiable Claims: 3/5
	2099	+ High Confidence Verdicts: 1
	2100	+ Medium Confidence: 2
	2101	+ Evidence Sources: 12
	2102	+ Avg Source Quality: 0.73
	2103	+ Quality Score: 8.5/10
	2104	+{{/code}}